CN117274005B - Big data pushing method and system based on digital education - Google Patents

Abstract

This application provides a big data push method and system based on digital education, by tracking the prior online learning events of target digital education users at preset online learning nodes and the first learning mode data tracked at the preset online learning nodes. Perform correlation analysis to generate event correlation analysis data, and then conduct feature inference and correlation configuration, which can accurately identify and infer students' learning behaviors and learning patterns, thereby generating high-quality teaching suggestions. Moreover, it takes into account the students' prior learning events and learning patterns, and can personalize the specific needs of each student, and then push the relevant configuration data into big data, which helps to optimize the digital education push strategy. Since students' learning behaviors and learning patterns can be dynamically analyzed, and inferences and configurations can be made based on credible values, it can dynamically adjust along with the students' learning process and provide the most appropriate teaching suggestions.

Description

Big data push method and system based on digital education

Technical field

This application relates to the field of digital education technology, specifically, to a big data push method and system based on digital education.

Background technique

With the rapid development of Internet technology, digital education has become an important part of the education field. Online learning platforms can provide convenient and flexible learning methods, allowing students to study at their own pace and needs. However, since each student has different learning habits, background knowledge, and understanding abilities, how to provide personalized teaching support according to the characteristics of each student has become an important issue facing current digital education.

At present, most online learning platforms mainly generate teaching suggestions by collecting and analyzing students' learning behavior data, such as learning time, learning progress, question answering, etc. However, this method often cannot accurately reflect students' learning patterns and cannot fully utilize a priori learning event information. In addition, due to the lack of effective correlation analysis and inference mechanisms, these systems are often unable to provide accurate and comprehensive teaching suggestions.

Therefore, developing a method that can deeply understand students' online learning behavior and thereby generate more accurate and personalized teaching suggestions has become an important research direction in the field of digital education.

Contents of the invention

In view of this, the purpose of this application is to provide a big data push method and system based on digital education.

According to the first aspect of this application, a big data push method based on digital education is provided and applied to a digital education system. The method includes:

Perform correlation analysis on multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and multiple first learning mode data tracked at the preset online learning node to generate event correlation sexual analysis data;

If it is determined based on the event correlation analysis data that one of the plurality of prior learning events has a target prior learning event that is not related to the first learning mode data, feature inference is performed based on the target prior learning event, and the The second learning mode data associated with the target prior learning event before the preset online learning node specifies the guess information corresponding to the learning node, and the guess information includes the second learning mode data corresponding to the speculative learning node. Speculative learning behavior and speculative learning representation parameters;

Based on the second learning mode data, the inferred learning representation parameters corresponding to the inferred learning node, the learning representation parameters corresponding to the terminal learning behavior in the target prior learning event, and the target in the target prior learning event The continuous parameters of the inferred learning behavior are determined to determine the inferred credible value of the inferred information; the target inferred learning behavior is used to characterize the inferred learning behavior after the target learning behavior that is finally triggered in the target prior learning event;

If the inferred trust value is not less than the set trust value, perform learning event correlation configuration on the inferred learning behavior corresponding to the inferred learning node of the second learning mode data and the target prior learning event, and Each correlation configuration data is pushed to the corresponding big data service system as the digital education big data of the target digital education user.

In a possible implementation of the first aspect, the inferred learning representation parameters corresponding to the inferred learning node and the terminal learning behavior in the target prior learning event based on the second learning mode data correspond to The learning representation parameters and the continuous parameters of the target inferred learning behavior in the target prior learning event are determined to determine the inferred credible value of the inferred information, including:

Determining the degree of deviation of the second learning mode data between the inferred learning representation parameter corresponding to the inferred learning node and the learning representation parameter corresponding to the terminal learning behavior in the target prior learning event;

Based on the target inferred continuous parameters of the learning behavior and the deviation degree, determine the inferred credibility value of the inferred information;

The presumed credibility value of the presumed information and the persistence parameter of the target presumed learning behavior are inversely correlated, and the presumed credible value of the presumed information and the deviation degree are inversely correlated.

In a possible implementation of the first aspect, the plurality of first learning mode data are collected through multiple online monitoring components;

If it is determined based on the event correlation analysis data that one of the plurality of prior learning events has a target prior learning event that is not related to the first learning mode data, feature inference is performed based on the target prior learning event, and it is determined The second learning mode data associated with the target prior learning event before the preset online learning node speculates the information corresponding to the learning node, including:

If the monitoring field intervals of the plurality of online monitoring components at the preset online learning node have shared field parts, and it is determined based on the event correlation analysis data that one of the plurality of prior learning events has the same content as the first learning mode Target a priori learning events with irrelevant data, perform feature inference based on the target a priori learning events, and determine a second learning mode data inference learning node associated with the target a priori learning event before the preset online learning node Corresponding speculation information.

In a possible implementation of the first aspect, the plurality of first learning mode data are collected based on multiple online monitoring components; the event correlation analysis data includes first correlation feature information;

The multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and the multiple first learning mode data tracked at the preset online learning node are subjected to correlation analysis to generate Event correlation analysis data, including:

Based on the monitoring service labels of the multiple online monitoring components for each of the a priori learning events and the monitoring service labels of the multiple online monitoring components for each of the first learning mode data, from the multiple a priori determining a reference learning event from the learning event and determining reference learning mode data from the plurality of first learning mode data;

Determine the reference learning event and the reference learning mode based on the monitoring service tags of the multiple online monitoring components for the reference learning event and the monitoring service tags of the multiple online monitoring components for the reference learning mode data. feature distance between data;

Based on the feature distance, learning event correlation configuration is performed on the reference learning event and the reference learning mode data, and the first correlation feature information is generated.

In a possible implementation of the first aspect, the monitoring service label based on the plurality of online monitoring components for each of the a priori learning events, and the plurality of online monitoring components for each of the first Monitoring service tags for learning mode data, determining reference learning events from the plurality of prior learning events and determining reference learning mode data from the plurality of first learning mode data, including:

Based on the monitoring service tags of the multiple online monitoring components for each of the a priori learning events, generate a monitoring service tag sequence corresponding to each of the a priori learning events;

If there is a target monitoring service tag sequence including the monitoring service tags of the multiple online monitoring components for the first learning mode data among the multiple monitoring service tag sequences, obtain the a priori learning event corresponding to the target monitoring service tag sequence. as a reference learning event, and obtain the first learning mode data as reference learning mode data.

In a possible implementation of the first aspect, the monitoring service tag based on the plurality of online monitoring components for the reference learning event and the monitoring of the reference learning mode data by the plurality of online monitoring components are The service tag determines the characteristic distance between the reference learning event and the reference learning pattern data, including:

Determine the weight value of the monitoring service tags of the multiple online monitoring components for the reference learning mode data in the monitoring service tag sequence corresponding to the reference learning event, as the reference learning mode data corresponding to the reference learning event. The service tag weight value;

The service tag weight value corresponding to the reference learning mode data under the reference learning event is compared with the matching degree between the reference learning event and the reference learning mode data to generate the reference learning event and the matching degree between the reference learning event and the reference learning mode data. Feature distance between the reference learning pattern data.

In a possible implementation of the first aspect, the event correlation analysis data further includes second correlation feature information;

The multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and the multiple first learning mode data tracked at the preset online learning node are subjected to correlation analysis to generate Event correlation analysis data also includes:

If based on the first correlation feature information, it is determined that among the plurality of prior learning events, there is a first learning event that is not related to the first learning mode data and that among the plurality of first learning mode data, there is a first learning event that is not related to the prior learning mode data. Learning event-independent third learning mode data, based on the matching degree between the first learning event and the third learning mode data, performing learning events on the first learning event and the third learning mode data Correlation configuration, generating second correlation feature information.

In a possible implementation of the first aspect, the event correlation analysis data further includes third correlation feature information;

The multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and the multiple first learning mode data tracked at the preset online learning node are subjected to correlation analysis to generate Event correlation analysis data also includes:

If based on the first correlation feature information and the second correlation feature information, it is determined that the plurality of prior learning events include a second learning event that is not related to the first learning mode data and the plurality of third learning events. One learning mode data has fourth learning mode data that is not related to the a priori learning event. For each fourth learning mode data, the fourth learning mode data is connected to the second learning mode that best matches the fourth learning mode data. On the event, third correlation feature information is generated;

If it is determined based on the first correlation feature information, the second correlation feature information and the third correlation feature information that the plurality of second learning mode data has fifth learning mode data that is not related to the a priori learning event, discard the first Five learning model data.

In a possible implementation of the first aspect, the plurality of first learning mode data are collected on an online course education page;

If based on the first correlation feature information and the second correlation feature information, it is determined that the plurality of prior learning events include a second learning event that is not related to the first learning mode data and the plurality of prior learning events. The first learning mode data includes fourth learning mode data that is not related to the a priori learning event. For each fourth learning mode data, the fourth learning mode data is connected to the fourth learning mode data that best matches the fourth learning mode data. On the second learning event, the third correlation feature information is generated, including:

If the online course education page is located in a learning mode sharing process, and based on the first correlation feature information and the second correlation feature information, it is determined that the plurality of prior learning events have data that is different from the first learning mode data. The related second learning event and the plurality of first learning mode data include fourth learning mode data that is not related to the a priori learning event. For each fourth learning mode data, the fourth learning mode data is associated with On the second learning event that the fourth learning mode data most matches, third correlation feature information is generated.

According to a second aspect of the present application, a digital education system is provided. The digital education system includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. The processor executes When the machine can execute instructions, the digital education system implements the aforementioned big data push method based on digital education.

According to a third aspect of the present application, a computer-readable storage medium is provided. Computer-executable instructions are stored in the computer-readable storage medium. When the computer-executable instructions are executed, the aforementioned digital-based education is realized. Big data push method.

According to any of the above aspects, in this application, correlation is first performed on multiple prior learning events of the target digital education user at the preset online learning node and multiple first learning mode data tracked at the preset online learning node. Analyze and generate event correlation analysis data. If it is determined based on the event correlation analysis data that there are target prior learning events that are not related to the first learning mode data, the system will further perform feature inference based on these target prior learning events to determine the third link before the preset online learning node. The second learning mode data infers the inference information corresponding to the learning node. The speculative information includes the speculative learning behavior and speculative learning representation parameters corresponding to the second learning mode data at the speculative learning node. Next, based on the second learning mode data, the inferred learning representation parameters corresponding to the inferred learning node, the learning representation parameters corresponding to the terminal learning behavior in the target prior learning event, and the continuation of the target inferred learning behavior in the target prior learning event are Parameters to determine the presumed confidence value of the speculated information. If the inferred trust value is not less than the set trust value, then the learning event correlation will be configured, the inferred learning behavior corresponding to the inferred learning node of the second learning mode data will be associated with the target prior learning event, and each related The sexual configuration data is used as digital education big data for target digital education users and then pushed to the corresponding big data service system. This will help generate more accurate personalized teaching suggestions, thereby optimizing the effect of online education.

That is, the embodiment of the present application generates event correlation by performing correlation analysis on the target digital education user's prior online learning events at the preset online learning node and the first learning mode data tracked at the preset online learning node. Analyzing data and then conducting feature inference and correlation configuration provides a more in-depth and comprehensive way to understand students' online learning behaviors, and can accurately identify and infer students' learning behaviors and learning patterns, thereby generating high-quality teaching. suggestion. Moreover, students' prior learning events and learning patterns are taken into account, so that the generated teaching suggestions can be personalized for each student's specific needs, and then the relevant configuration data is pushed into big data, which helps optimize digital education push. Strategy. Since students' learning behaviors and learning patterns can be dynamically analyzed, and inferences and configurations can be made based on credible values, it can dynamically adjust along with the students' learning process and provide the most appropriate teaching suggestions.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application and therefore do not It should be regarded as a limitation of the scope. For those of ordinary skill in the art, other relevant drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is a schematic flow chart of the big data push method based on digital education provided by the embodiment of this application;

Figure 2 shows a schematic component structure diagram of a digital education system provided by an embodiment of the present application for implementing the above-mentioned big data push method based on digital education.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below based on the drawings in the embodiments of the present application. It should be understood that the technical solutions attached in the embodiments of the present application The drawings are for illustration and description purposes only and are not intended to limit the scope of the present application. Additionally, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented in accordance with some embodiments of the embodiments of this application. It should be understood that the operations of the flowchart may be implemented out of sequence, and steps without logical context may be implemented in reverse order or simultaneously. In addition, those skilled in the art can add at least one other operation to the flow chart and destroy at least one operation from the flow chart under the guidance of the content of this application.

In addition, the described embodiments are only some of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations. Accordingly, the following detailed description of the embodiments of the application provided in the appended drawings is not intended to limit the scope of the claimed application, but rather to represent selected embodiments of the application. According to the embodiments of the present application, all other embodiments obtained by those skilled in the art without any creative work shall correspond to the protection scope of the present application.

Figure 1 shows a schematic flow chart of the big data push method and system based on digital education provided by the embodiment of the present application. It should be understood that in other embodiments, some steps of the big data push method based on digital education in this embodiment are as follows: Sequences can be shared with each other based on actual needs, or some steps can be omitted or maintained. The detailed steps of this big data push method based on digital education include:

Step S110, perform correlation analysis on the multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and the multiple first learning mode data tracked at the preset online learning node, Generate event correlation analysis data.

For example, suppose there is an online physics course with preset online learning nodes covering topics such as mechanics, electromagnetism, and quantum physics. Student A has completed the study of mechanics and electromagnetism, and has entered the learning node of quantum physics. At this time, student A’s behavior patterns in previous learning nodes (such as mechanics and electromagnetism) can be recorded, such as the time they spent, the number of exercises completed, error types, etc., and compared with the first learning node of quantum physics Pattern data (which may come from the learning behaviors of other users in quantum physics nodes) are subjected to correlation analysis to generate event correlation analysis data.

That is, in the above example of the physics course, the target digital education user refers to student A, who is a student who hopes to provide personalized teaching suggestions by analyzing his learning behavior and patterns.

The preset online learning nodes refer to various subject parts of the online physics course, such as mechanics, electromagnetism, and quantum physics. Each topic section can be viewed as a learning node.

The multiple prior learning events of the prior online learning node refer to the learning behavior of student A before completing quantum physics, that is, on the two online learning nodes of mechanics and electromagnetism. These learning behaviors may include the time he spends solving problems, the speed with which he completes assignments, the correctness of answering questions, etc.

The multiple first learning mode data tracked by the preset online learning node refers to the learning behavior data collected from other students at the quantum physics node, such as their average time to solve problems, error types, etc.

Conduct correlation analysis on Student A's learning events at the mechanics and electromagnetics nodes and the learning pattern data tracked at the quantum physics node to generate event correlation analysis data, which means that student A's learning behavior at the mechanics and electromagnetics nodes can be compared with Check the learning behavior of other students at quantum physics nodes to see if there is any correlation. If Student A’s learning behavior at the mechanics and electromagnetism nodes is significantly different from the successful learning behavior of other students at the quantum physics node, it will be considered that student A may encounter difficulties at the quantum physics node, and subsequent speculations and predictions will be made accordingly. suggestion.

Step S120, if it is determined based on the event correlation analysis data that one of the multiple prior learning events has a target prior learning event that is not related to the first learning mode data, perform feature inference based on the target prior learning event, Determine the second learning mode data associated with the target prior learning event before the preset online learning node and the speculation information corresponding to the speculation learning node. The speculation information includes the second learning mode data in the speculation learning node. The speculative learning behavior and speculative learning representation parameters corresponding to the node.

For example, if the analysis finds that certain learning behaviors (target prior learning events) of student A at the mechanics node are not correlated with the learning pattern of the quantum physics node, then feature inference can be made on these target prior learning events. For example, if it is found that student A is too focused on the application of formulas at the mechanics node and neglects the understanding of theoretical knowledge, and needs a deeper theoretical understanding at the quantum physics node, then it can be inferred that student A may encounter difficulties at the quantum physics node. .

That is, in the example of this physics course, the event correlation analysis data refers to the data obtained by comparing the learning behavior of student A on the mechanics and electromagnetism nodes with the learning behavior of other students on the quantum physics node.

If certain learning behaviors of Student A at the nodes of mechanics and electromagnetism (such as over-reliance on formula application and neglect of theoretical understanding) are different from the learning patterns of other students who successfully completed quantum physics learning, then these prior learning events are considered Targeted prior learning events that are not related to the first learning mode data.

At this time, feature inference can be performed based on these target prior learning events. In other words, you can try to predict what kind of learning behavior student A may show in the quantum physics node if he continues his learning behaviors in the mechanics and electromagnetics nodes (such as over-reliance on formula application).

The speculative information includes the predicted possible learning behavior of student A at the quantum physics node (for example, he may encounter difficulties in understanding theoretical knowledge) and the representation parameters of this prediction (such as the possible degree of difficulty, the possible impact on learning progress, etc.). This speculation information will be used for subsequent determination of the presumed credible value and learning event correlation configuration.

Step S130, based on the second learning mode data, the inferred learning representation parameters corresponding to the inferred learning node, the learning representation parameters corresponding to the terminal learning behavior in the target a priori learning event, and the target a priori learning event The continuous parameters of the target inferred learning behavior are determined to determine the inferred credible value of the inferred information. The target speculative learning behavior is used to characterize the speculative learning behavior after the target learning behavior finally triggered in the target prior learning event.

For example, the credibility of this speculation can be determined based on Student A's last learning behavior at the mechanics node, the inferred learning behavior at the quantum physics node, and the possible duration of this inferred behavior. For example, if Student A's behavior of paying too much attention to the application of formulas at the mechanics node continues until the end, then the speculation that he encountered difficulties at the quantum physics node will be more credible.

That is to say, in the example of this physics course, the speculative learning representation parameters corresponding to the second learning mode data at the speculative learning node refer to the predicted learning behavior that Student A may occur at the quantum physics node and this learning behavior. Characteristics. For example, it might be predicted that Student A will have difficulty understanding quantum physics theory due to an overreliance on formula application.

The learning representation parameters corresponding to the terminal learning behavior in the target prior learning event refer to the learning behavior of student A at the end of the previous learning node (electromagnetism) and the characteristics of this learning behavior. For example, if Student A still relies too much on formula application at the end of the electromagnetics section, then this learning behavior is a terminal learning behavior. The continuous parameters of the target inferred learning behavior may include the time and frequency of Student A's continued over-reliance on the application of formulas, etc.

The credibility value of the speculative information can be determined based on the above three parameters, that is, the credibility of predicting that Student A may encounter difficulties at the quantum physics node is evaluated. If the above three parameters show that student A is likely to encounter difficulties in understanding quantum physics theory, then the credibility value of the speculation will be very high.

Target inferred learning behavior refers to student A’s possible learning behavior at the quantum physics node (encountering difficulties in understanding theoretical knowledge) predicted based on his terminal learning behavior at the electromagnetics node (over-reliance on formula application). This goal speculates that learned behaviors will be used in subsequent learned event correlation configurations.

Step S140: If the inferred trust value is not less than the set trust value, perform learning event correlation on the inferred learning behavior corresponding to the inferred learning node of the second learning mode data and the target prior learning event. Configuration, and each correlation configuration data is pushed to the corresponding big data service system as the digital education big data of the target digital education user.

For example, if the inferred credibility value is higher than a certain threshold, then the system will associate student A’s learning behavior at the mechanics node with the inferred learning behavior at the quantum physics node, and use this information as student A’s digital education Big data is sent to the big data service system. For example, if the credibility of the speculation is very high, the system can remind student A to strengthen the study of theoretical knowledge before starting the quantum physics node, or recommend him some additional resources to deepen theoretical understanding.

That is, in the example of the physics course, the conjecture credibility value refers to the credibility of the prediction that Student A may encounter difficulties at the quantum physics node. This credibility is calculated based on parameters such as Student A's learning behavior at the end of the electromagnetics node, the predicted learning behavior of the quantum physics node, and the possible duration of this behavior. If the estimated credibility value is not less than the set threshold (such as 0.7 or 70%), it means that the current prediction result is relatively accurate. At this point, the next step can be taken, which is to correlate student A’s learning behavior at the electromagnetics node (such as over-reliance on formula application) with the predicted learning behavior at the quantum physics node (such as possible difficulty in understanding theoretical knowledge) . In other words, record this information: Student A may encounter difficulties in understanding theoretical knowledge when learning quantum physics due to over-reliance on formula application. Then, these associated data are sent to the big data service system as student A's digital education big data. There, these digital education data can be further analyzed and used to generate more personalized learning suggestions, such as reminding student A to strengthen the study of theoretical knowledge before starting a quantum physics node, or recommending him some additional resources to deepen his theoretical understanding. . In this way, student A can be better helped to study effectively.

Based on the above steps, first perform correlation analysis on multiple prior learning events of the target digital education user at the preset online learning node and multiple first learning mode data tracked at the preset online learning node to generate event correlation. analyze data. If it is determined based on the event correlation analysis data that there are target prior learning events that are not related to the first learning mode data, the system will further perform feature inference based on these target prior learning events to determine the third link before the preset online learning node. The second learning mode data infers the inference information corresponding to the learning node. The speculative information includes the speculative learning behavior and speculative learning representation parameters corresponding to the second learning mode data at the speculative learning node. Next, based on the second learning mode data, the inferred learning representation parameters corresponding to the inferred learning node, the learning representation parameters corresponding to the terminal learning behavior in the target prior learning event, and the continuation of the target inferred learning behavior in the target prior learning event are Parameters to determine the presumed confidence value of the speculated information. If the inferred trust value is not less than the set trust value, then the learning event correlation will be configured, the inferred learning behavior corresponding to the inferred learning node of the second learning mode data will be associated with the target prior learning event, and each related The sexual configuration data is used as digital education big data for target digital education users and then pushed to the corresponding big data service system. This will help generate more accurate personalized teaching suggestions, thereby optimizing the effect of online education.

That is, the embodiment of the present application generates event correlation by performing correlation analysis on the target digital education user's prior online learning events at the preset online learning node and the first learning mode data tracked at the preset online learning node. Analyzing data and then conducting feature inference and correlation configuration provides a more in-depth and comprehensive way to understand students' online learning behaviors, and can accurately identify and infer students' learning behaviors and learning patterns, thereby generating high-quality teaching. suggestion. Moreover, students' prior learning events and learning patterns are taken into account, so that the generated teaching suggestions can be personalized for each student's specific needs, and then the relevant configuration data is pushed into big data, which helps optimize digital education push. Strategy. Since students' learning behaviors and learning patterns can be dynamically analyzed, and inferences and configurations can be made based on credible values, it can dynamically adjust along with the students' learning process and provide the most appropriate teaching suggestions.

In a possible implementation, step S130 may include:

Step S131: Determine the degree of deviation of the second learning mode data between the inferred learning representation parameter corresponding to the inferred learning node and the learning representation parameter corresponding to the terminal learning behavior in the target prior learning event.

For example, in the physics course example, one could identify predicted learning behaviors (e.g., student A might have difficulty understanding quantum physics theory) versus his actual learning behaviors at the end of the electromagnetics node (e.g., overreliance on formula application ). If the difference between the two is large, the degree of deviation will be high; conversely, if the two are similar, the degree of deviation will be low.

For example, Euclidean distance can be used to measure the predicted learning representation parameters (i.e., the predicted degree to which Student A may encounter difficulties at the quantum physics node) and the learning representation parameters corresponding to the terminal learning behavior (i.e., Student A's predicted difficulty in electromagnetic physics). the degree of overreliance on formula application at the end of the learning node). Assuming that both are quantified as values from 0 to 10, where 10 represents the highest degree, then the degree of deviation can be calculated as the Euclidean distance between them, which is the absolute value | Predictive learning representation parameters - terminal learning behavior correspondence The learned representation parameters of |.

Step S132: Based on the persistence parameters of the target inference learning behavior and the deviation degree, determine the inference credibility value of the inference information. Wherein, the presumed credibility value of the presumed information and the persistence parameter of the target presumed learning behavior are inversely correlated, and the presumed credible value of the presumed information and the deviation degree are inversely correlated.

For example, after determining the degree of deviation, it is also necessary to consider how long Student A's over-reliance on formula application may continue. If he is still overly reliant on formula application at the end of the Electromagnetics node, then the persistence parameter for this behavior will be high. Combining these two factors, the credible value of the forecast information can be determined.

Among them, if the behavior of student A that relies too much on formula application lasts for a long time (that is, the persistence parameter is high), or the gap between the predicted learning behavior and the actual learning behavior is large (that is, the degree of deviation is high), then for predicting student A’s performance in quantum physics The credibility that a node may encounter difficulties is reduced. That is, the higher the persistence parameters and deviation, the lower the confidence value of the prediction. This is because if Student A’s learning behavior is persistent, or his learning behavior is very different from the predicted learning behavior, then the possibility of him changing his learning behavior will decrease, resulting in increased prediction inaccuracy.

For example, the guessed trust value can be defined as a number between 0 and 1, where 1 represents completely credible and 0 represents completely untrustworthy. Assuming that the persistence parameter of the target inferred learning behavior (that is, the length of time that Student A's behavior of overly relying on the application of the formula may last) is also quantified as a value from 0 to 10, then the inferred credible value can be calculated as 1 - (0.5 * deviation degree/10 + 0.5 * duration parameter/10). This formula assumes that the deviation and persistence parameters have an equal impact on the estimated credible value, and that they have a negative impact, that is, the larger the value, the smaller the credible value.

The above is just a simplified example, and actual calculations may involve more complex statistical methods, such as using multiple linear regression, logistic regression, or machine learning models to consider the interaction and nonlinear relationships of various factors. But no matter what, the goal of these calculations is the same: to assess the credibility of predictions of future learning behavior based on available information.

For example, in some other examples, if you consider more parameter factors, for example, you may want to add a parameter that reflects the student's overall performance at the previous node (such as the average score M), and a parameter that reflects the student's learning enthusiasm or participation. parameters (such as the activity level L on the online learning platform). Then the formula can be rewritten as:

Assuming that P represents the "presumed learning representation parameter" and A represents the "learning representation parameter corresponding to the final learning behavior", then the deviation D can be calculated as:

D = |P - A| / (1 + M/100)

It is assumed here that the average score M is a value from 0 to 100, and it has a negative impact on the deviation: the higher the score, the smaller the deviation.

Assuming that T represents the "continuous parameter of the target inferred learning behavior", then the inferred credible value C can be calculated as:

C = 1 - 0.5 * (D/10 + T/10) + L/100

It is assumed here that the activity level L is a value from 0 to 100, and it has a positive impact on the inferred credibility value: the higher the activity level, the greater the credibility value.

The above formula is an extended example, and more complex modeling may be required based on actual conditions and data in actual applications.

In a possible implementation, the plurality of first learning mode data are collected through multiple online monitoring components.

If it is determined based on the event correlation analysis data that one of the plurality of prior learning events has a target prior learning event that is not related to the first learning mode data, feature inference is performed based on the target prior learning event, and it is determined The second learning mode data associated with the target prior learning event before the preset online learning node speculates the information corresponding to the learning node, including:

If the monitoring field intervals of the plurality of online monitoring components at the preset online learning node have shared field parts, and it is determined based on the event correlation analysis data that one of the plurality of prior learning events has the same content as the first learning mode Target a priori learning events with irrelevant data, perform feature inference based on the target a priori learning events, and determine a second learning mode data inference learning node associated with the target a priori learning event before the preset online learning node Corresponding speculation information.

In the physics course example, the online monitoring component can be various functions of the online education platform, such as students' test scores, study time, frequency of participation in discussions, etc. These online monitoring components will collect student A’s learning behavior data on mechanics and electromagnetics nodes, as well as other students’ learning behavior data on quantum physics nodes.

Assume that the online monitoring component collects data on student A's test scores, study time, etc. at the mechanics and electromagnetics nodes, and collects the same data on other students at the quantum physics node. This means that the monitored field intervals of these online monitoring components (i.e., the types of data they collect) have shared field parts (i.e., test scores, study time, etc.). Then, you can compare the learning behavior of student A on the mechanics and electromagnetism nodes with the learning behavior of other students on the quantum physics node. If it is found that there is a significant difference between the learning behavior of student A and the behavior of students who successfully completed quantum physics learning, then These learning behaviors are considered as target prior learning events that are not related to the first learning mode data. Then, feature inference is performed based on these prior learning events to predict the possible learning behavior of student A at the quantum physics node.

In a possible implementation, the plurality of first learning mode data are collected based on multiple online monitoring components. The event correlation analysis data includes first correlation feature information.

For example, in the physics course example, the online monitoring component may be an online education platform that can track and record the learning behavior data of student A on mechanics and electromagnetics nodes, as well as the learning behavior data of other students on quantum physics nodes. This data includes study time, number of exercises completed, error types, etc.

Step S110 may include:

Step S111: Based on the monitoring service tags of the multiple online monitoring components for each of the a priori learning events and the monitoring service tags of the multiple online monitoring components for each of the first learning mode data, from the multiple online monitoring components A reference learning event is determined from a priori learning events and reference learning mode data is determined from the plurality of first learning mode data.

For example, reference learning events and reference learning pattern data may be determined based on data collected by the online monitoring component. For example, the reference learning event may be student A's behavior of over-reliance on formula application on the electromagnetics node, while the reference learning mode data may be the learning behavior of other students on the quantum physics node.

Step S112: Based on the monitoring service tags of the multiple online monitoring components for the reference learning event and the monitoring service tags of the multiple online monitoring components for the reference learning mode data, determine the relationship between the reference learning event and the Feature distance between reference learning pattern data.

For example, the feature distance between the reference learning event and the reference learning pattern data can be calculated based on their differences in different features. For example, if student A relies too much on formula application in the electromagnetics node, while other students focus more on theoretical understanding in the quantum physics node, then the characteristic distance between the two may be larger.

Step S113: Based on the feature distance, perform learning event correlation configuration on the reference learning event and the reference learning mode data, and generate the first correlation feature information.

For example, the correlation between the reference learning event and the reference learning pattern data can be analyzed based on the calculated feature distance, and correlation feature information can be generated. For example, if the feature distance is large, indicating that there is a significant difference between the learning behavior of student A at the electromagnetics node and the learning behavior of other students at the quantum physics node, then the correlation feature information may show that the correlation between the two is low. .

In a possible implementation, step S111 may include:

Step S1111: Based on the monitoring service labels of the multiple online monitoring components for each of the a priori learning events, generate a monitoring service label sequence corresponding to each of the a priori learning events.

For example, a monitoring service tag sequence for each a priori learning event can be generated based on the data collected by the online monitoring component. For example, for each learning behavior of student A on the mechanics and electromagnetics nodes (such as the number of exercises completed, error types, etc.), a corresponding monitoring service label will be generated, and these monitoring service labels will be composed into a certain order. sequence.

Step S1112: If there is a target monitoring service tag sequence including the monitoring service tags of the multiple online monitoring components for the first learning mode data among the multiple monitoring service tag sequences, obtain the prior information corresponding to the target monitoring service tag sequence. The experimental learning event is used as a reference learning event, and the first learning mode data is obtained as the reference learning mode data.

For example, you can search for those monitoring service tag sequences that include the monitoring service tags of the online monitoring component for the first learning mode data (that is, the learning behavior data of other students on the quantum physics node). For example, if there is a label sequence that represents student A's behavior of over-reliance on formula application in the electromagnetics node, and this behavior also exists in the learning behavior of other students in the quantum physics node, then this label sequence will be selected. Target monitoring service tag sequence, corresponding learning events (over-reliance on formula application) and learning pattern data (learning behaviors of other students on quantum physics nodes) will be selected as reference learning events and reference learning pattern data.

In a possible implementation, step S112 may include:

Step S1121, determine the weight value of the monitoring service tags of the multiple online monitoring components for the reference learning mode data in the monitoring service tag sequence corresponding to the reference learning event, as the reference learning under the reference learning event. The service tag weight value corresponding to the pattern data.

For example, the monitoring service tag of the online monitoring component for the reference learning mode data (i.e., the learning behavior of other students on the quantum physics node) can be calculated to correspond to the reference learning event (such as student A's over-reliance on formula application on the electromagnetics node). The service label weight value in the monitoring service label sequence. This service tag weight value reflects the importance of the reference learning model data in the reference learning event.

Step S1122, determine the ratio between the service tag weight value corresponding to the reference learning mode data under the reference learning event and the matching degree between the reference learning event and the reference learning mode data, and generate the reference The characteristic distance between the learning event and the reference learning pattern data.

For example, the matching degree between the reference learning event and the reference learning mode data can be calculated, and then the service tag weight value is compared with the matching degree to obtain the feature distance. This feature distance reflects the similarity or difference between the reference learning event and the reference learning pattern data. For example, if the feature distance is large, it may indicate that there is a significant difference between student A's learning behavior on the electromagnetics node and other students' learning behavior on the quantum physics node.

In a possible implementation, the event correlation analysis data further includes second correlation feature information.

Step S110 may also include: based on the first correlation feature information, it is determined that among the plurality of prior learning events, there is a first learning event that is not related to the first learning mode data and the plurality of first learning events. The pattern data includes third learning pattern data that is not related to the a priori learning event. Based on the matching degree between the first learning event and the third learning pattern data, the first learning event and the third learning pattern data are compared. The three learning mode data performs correlation configuration of learning events to generate second correlation feature information.

For example, based on the calculated first correlation feature information, it is possible to find those learning events that are not related to the first learning mode data (that is, the learning behavior of other students on the quantum physics node), and those learning events that are not related to the prior learning events. (That is, student A’s learning behavior on the mechanics and electromagnetics nodes) Irrelevant learning mode data. For example, if Student A spends a lot of time reading theory on the Mechanics node but relies too much on formula application on the Electromagnetics node, then these two behaviors may be viewed as unrelated learning events. Then, the matching degree between these irrelevant learning events and the learning mode data is calculated, and correlation configuration is performed based on this matching degree, thereby generating second correlation feature information. For example, if the time Student A spent reading theory on the Mechanics node closely matches the time another student spent reading theory on the Quantum Physics node, then a correlation configuration between the two might yield a higher correlation. characteristic value as the second correlation characteristic information.

In a possible implementation, the event correlation analysis data further includes third correlation feature information.

Step S110 may also include: based on the first correlation feature information and the second correlation feature information, it is determined that the plurality of prior learning events include second learning that is not relevant to the first learning mode data. The event and the plurality of first learning mode data include fourth learning mode data that is not related to the a priori learning event. For each fourth learning mode data, the fourth learning mode data is associated with the fourth learning mode. On the second learning event that the data best matches, the third correlation feature information is generated. If it is determined based on the first correlation feature information, the second correlation feature information and the third correlation feature information that the plurality of second learning mode data has fifth learning mode data that is not related to the a priori learning event, discard the first Five learning model data.

For example, based on the calculated first and second correlation feature information, it is possible to find out those data that are related to the first learning mode data (i.e., the learning behavior of other students on the quantum physics node) and the prior learning events (i.e., student A’s Learning events and learning pattern data that are not related to learning behaviors on mechanics and electromagnetics nodes). Then, for each of the found irrelevant fourth learning mode data, find the second learning event that best matches it, and correlate them to generate third correlation feature information. Then, based on all the correlation feature information that has been calculated, find out those fifth learning mode data that are not relevant to the prior learning events, and discard these data. This means that this discarded data is no longer considered in subsequent analysis and recommendation generation.

In a possible implementation, the plurality of first learning mode data are collected on an online course education page.

Then in the above embodiment, if the online course education page is located in the learning mode sharing process, and based on the first correlation feature information and the second correlation feature information, it is determined that the multiple prior learning events have The second learning event that is not related to the first learning mode data and the fourth learning mode data that is not related to the a priori learning event among the plurality of first learning mode data, for each fourth learning mode data, the third learning mode data is The fourth learning mode data is linked to the second learning event that best matches the fourth learning mode data to generate third correlation feature information.

For example, first confirm that the collected first learning mode data (that is, the learning behavior of other students on quantum physics nodes) is obtained through the online course education page. For example, the data might come from an online education platform that provides physics courses and records student behavior during the learning process. If the online course education page is located in a shared learning mode process, then the system will find out those data that are related to the first learning mode data (i.e. other students on the quantum physics node) based on the calculated first and second correlation feature information. learning behavior) and prior learning events (that is, student A’s learning behavior on mechanics and electromagnetics nodes), and associate them to generate third correlation feature information.

This shared learning model process can be understood as the backend service of an online education platform, which is responsible for managing and analyzing the learning model data of all online courses. In this way, the system can analyze learning behaviors on a wider scale, thereby providing more comprehensive and precise personalized teaching suggestions.

Figure 2 schematically illustrates a digital education system 100 that may be used to implement various embodiments described in this application.

For one embodiment, Figure 2 illustrates a digital education system 100 having at least one processor 102, a control module (chipset) 104 coupled to at least one of the (at least one) processors 102, Memory 106 coupled to control module 104 , non-volatile memory (NVY)/storage device 108 coupled to control module 104 , at least one input/output device 110 coupled to control module 104 , and control Network interface 112 of module 104.

The processor 102 may include at least one single-core or multi-core processor, and the processor 102 may include any combination of a general-purpose processor or a special-purpose processor (eg, a graphics processor, an application processor, a baseband processor, etc.). In an alternative implementation, the digital education system 100 can serve as a server device such as the gateway described in the embodiment of this application.

Figure 2 schematically illustrates a digital education system 100 that may be used to implement various embodiments described in this application.

For one embodiment, Figure 2 illustrates a digital education system 100 having at least one processor 102, a control module (chipset) 104 coupled to at least one of the (at least one) processors 102, Memory 106 coupled to control module 104 , non-volatile memory (NVM)/storage device 108 coupled to control module 104 , at least one input/output device 110 coupled to control module 104 , and control Network interface 112 of module 104.

The processor 102 may include at least one single-core or multi-core processor, and the processor 102 may include any combination of a general-purpose processor or a special-purpose processor (eg, a graphics processor, an application processor, a baseband processor, etc.). In an alternative implementation, the digital education system 100 can serve as a server device such as the gateway described in the embodiment of this application.

In an alternative implementation, digital education system 100 may include at least one computer-readable medium (eg, memory 106 or NVM/storage device 108) having instructions 114 and configured in conjunction with the at least one computer-readable medium. At least one processor 102 executes instructions 114 to implement modules for performing the actions described in this disclosure.

For one embodiment, the control module 104 may include any suitable interface controller to provide any suitable interface to at least one of the (at least one) processors 102 and/or any suitable device or component in communication with the control module 104 .

Control module 104 may include a memory controller module to provide an interface to memory 106 . The memory controller module may be a hardware module, a software module, and/or a firmware module.

Memory 106 may be used, for example, to load and store data and/or instructions 114 for digital education system 100 . For one embodiment, memory 106 may include any suitable volatile memory, such as suitable DRAM. In an alternative implementation, memory 106 may include double data rate type quad synchronous dynamic random access memory (DDR4SDRAM).

For one embodiment, the control module 104 may include at least one input/output controller to provide an interface to the NVM/storage device 108 and (at least one) input/output device 110 .

For example, NVM/storage device 108 may be used to store data and/or instructions 114 . NVM/storage device 108 may include any suitable non-volatile memory (eg, flash memory) and/or may include any suitable (at least one) non-volatile storage device (eg, at least one hard disk drive (HDD), at least a compact disc (CD) drive and/or at least one digital versatile disc (DVD) drive).

NVM/storage device 108 may include storage resources that are physically part of the device on which digital education system 100 is installed, or may be accessible to the device without necessarily being part of the device. For example, NVM/storage device 108 may be accessed via (at least one) input/output device 110 over the network.

(At least one) input/output device 110 may provide an interface for the digital education system 100 to communicate with any other appropriate device. The input/output device 110 may include a communication component, a pinyin component, an online monitoring component, and the like. The network interface 112 may provide an interface for the digital education system 100 to communicate in accordance with at least one network, and the digital education system 100 may communicate with at least one component of the wireless network in accordance with any of at least one wireless network standard and/or protocol. Wireless communication, such as access to a wireless network based on communication standards, or a combination thereof for wireless communication.

For one embodiment, at least one of the (at least one) processors 102 may be loaded with logic for at least one controller (eg, a memory controller module) of the control module 104 . For one embodiment, at least one of the processor(s) 102 may be loaded together with the logic of at least one controller of the control module 104 to form a system level load. For one embodiment, at least one of the processor(s) 102 may be integrated on the same die as the logic of at least one controller of the control module 104 . For one embodiment, at least one of the (at least one) processors 102 may be integrated on the same die with the logic of at least one controller of the control module 104 to form a system on a chip (SoC).

In various embodiments, the digital education system 100 may be, but is not limited to: a server, a desktop computing device or a mobile computing device (eg, a laptop computing device, a handheld computing device, a tablet, a netbook, etc.) and other terminal devices. In various embodiments, digital education system 100 may have more or fewer components and/or a different architecture. For example, in an alternative implementation, the digital education system 100 includes at least one camera, a keyboard, a liquid crystal display (LCD) screen (including a touch screen display), a non-volatile memory port, multiple antennas, a graphics chip, a dedicated Integrated circuits (ASICs) and speakers.

The embodiments of the present application have been introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and the core idea of the present application; at the same time, for Those of ordinary skill in the art will have changes in the specific implementation and application scope based on the ideas of the present application. In summary, the content of this description should not be understood as a limitation of the present application.

Claims (9)

1. A big data push method based on digital education, characterized in that it is applied to a digital education system, and the method includes: Perform correlation analysis on multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and multiple first learning mode data tracked at the preset online learning node to generate event correlation Sexual analysis data, the first learning mode data refers to learning behavior data at preset online learning nodes collected from other digital education users; If it is determined based on the event correlation analysis data that one of the plurality of prior learning events has a target prior learning event that is not related to the first learning mode data, feature inference is performed based on the target prior learning event, and the The second learning mode data associated with the target prior learning event before the preset online learning node specifies the guess information corresponding to the learning node, and the guess information includes the second learning mode data corresponding to the speculative learning node. Speculative learning behavior and speculative learning representation parameters; Based on the second learning mode data, the inferred learning representation parameters corresponding to the inferred learning node, the learning representation parameters corresponding to the terminal learning behavior in the target prior learning event, and the target in the target prior learning event The continuous parameters of the inferred learning behavior are determined to determine the inferred credible value of the inferred information; the target inferred learning behavior is used to characterize the inferred learning behavior after the target learning behavior that is finally triggered in the target prior learning event; If the inferred trust value is not less than the set trust value, perform learning event correlation configuration on the inferred learning behavior corresponding to the inferred learning node of the second learning mode data and the target prior learning event, and Push each correlation configuration data to the corresponding big data service system as the digital education big data of the target digital education user; The second learning mode data is based on the speculative learning representation parameters corresponding to the speculative learning node, the learning representation parameters corresponding to the terminal learning behavior in the target prior learning event, and the target prior learning event. The goal is to infer the continuous parameters of the learning behavior and determine the inferred credible value of the inferred information, including: Determining the degree of deviation of the second learning mode data between the inferred learning representation parameter corresponding to the inferred learning node and the learning representation parameter corresponding to the terminal learning behavior in the target prior learning event; Based on the target inferred continuous parameters of the learning behavior and the deviation degree, determine the inferred credibility value of the inferred information; The presumed credibility value of the presumed information and the persistence parameter of the target presumed learning behavior are inversely correlated, and the presumed credible value of the presumed information and the deviation degree are inversely correlated. 2. The big data push method based on digital education according to claim 1, characterized in that the plurality of first learning mode data are collected through multiple online monitoring components; If it is determined based on the event correlation analysis data that one of the plurality of prior learning events has a target prior learning event that is not related to the first learning mode data, feature inference is performed based on the target prior learning event, and it is determined The second learning mode data associated with the target prior learning event before the preset online learning node speculates the information corresponding to the learning node, including: If the monitoring field intervals of the plurality of online monitoring components at the preset online learning node have shared field parts, and it is determined based on the event correlation analysis data that one of the plurality of prior learning events has the same content as the first learning mode Target a priori learning events with irrelevant data, perform feature inference based on the target a priori learning events, and determine a second learning mode data inference learning node associated with the target a priori learning event before the preset online learning node Corresponding speculation information. 3. The big data push method based on digital education according to claim 1, characterized in that the plurality of first learning mode data are collected based on a plurality of online monitoring components; the event correlation analysis data includes the first 1. Relevance feature information; The multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and the multiple first learning mode data tracked at the preset online learning node are subjected to correlation analysis to generate Event correlation analysis data, including: Based on the monitoring service labels of the multiple online monitoring components for each of the a priori learning events and the monitoring service labels of the multiple online monitoring components for each of the first learning mode data, from the multiple a priori determining a reference learning event from the learning event and determining reference learning mode data from the plurality of first learning mode data; Determine the reference learning event and the reference learning mode based on the monitoring service tags of the multiple online monitoring components for the reference learning event and the monitoring service tags of the multiple online monitoring components for the reference learning mode data. feature distance between data; Based on the feature distance, learning event correlation configuration is performed on the reference learning event and the reference learning mode data, and the first correlation feature information is generated. 4. The big data push method based on digital education according to claim 3, characterized in that the monitoring service tags for each of the a priori learning events based on the plurality of online monitoring components, and the plurality of The online monitoring component determines a reference learning event from the plurality of prior learning events and the reference learning mode data from the plurality of first learning mode data for the monitoring service tag of each of the first learning mode data, including : Based on the monitoring service tags of the multiple online monitoring components for each of the a priori learning events, generate a monitoring service tag sequence corresponding to each of the a priori learning events; If there is a target monitoring service tag sequence including the monitoring service tags of the multiple online monitoring components for the first learning mode data among the multiple monitoring service tag sequences, obtain the a priori learning event corresponding to the target monitoring service tag sequence. as a reference learning event, and obtain the first learning mode data as reference learning mode data. 5. The big data push method based on digital education according to claim 3, characterized in that the monitoring service tag based on the multiple online monitoring components for the reference learning event and the multiple online monitoring components Determining the characteristic distance between the reference learning event and the reference learning mode data for the monitoring service tag of the reference learning mode data includes: Determine the weight value of the monitoring service tags of the multiple online monitoring components for the reference learning mode data in the monitoring service tag sequence corresponding to the reference learning event, as the reference learning mode data corresponding to the reference learning event. The service tag weight value; The service tag weight value corresponding to the reference learning mode data under the reference learning event is compared with the matching degree between the reference learning event and the reference learning mode data to generate the reference learning event and the matching degree between the reference learning event and the reference learning mode data. Feature distance between the reference learning pattern data. 6. The big data push method based on digital education according to claim 3, characterized in that the event correlation analysis data also includes second correlation feature information; The multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and the multiple first learning mode data tracked at the preset online learning node are subjected to correlation analysis to generate Event correlation analysis data also includes: If based on the first correlation feature information, it is determined that among the plurality of prior learning events, there is a first learning event that is not related to the first learning mode data and that among the plurality of first learning mode data, there is a first learning event that is not related to the prior learning mode data. Learning event-independent third learning mode data, based on the matching degree between the first learning event and the third learning mode data, performing learning events on the first learning event and the third learning mode data Correlation configuration, generating second correlation feature information. 7. The big data push method based on digital education according to claim 6, characterized in that the event correlation analysis data also includes third correlation feature information; The multiple prior learning events of the target digital education user at the prior online learning node of the preset online learning node and the multiple first learning mode data tracked at the preset online learning node are subjected to correlation analysis to generate Event correlation analysis data also includes: If based on the first correlation feature information and the second correlation feature information, it is determined that the plurality of prior learning events include a second learning event that is not related to the first learning mode data and the plurality of third learning events. One learning mode data has fourth learning mode data that is not related to the a priori learning event. For each fourth learning mode data, the fourth learning mode data is connected to the second learning mode that best matches the fourth learning mode data. On the event, third correlation feature information is generated; If it is determined based on the first correlation feature information, the second correlation feature information and the third correlation feature information that the plurality of second learning mode data has fifth learning mode data that is not related to the a priori learning event, discard the first Five learning model data. 8. The big data push method based on digital education according to claim 7, characterized in that the plurality of first learning mode data are collected on the online course education page; If based on the first correlation feature information and the second correlation feature information, it is determined that the plurality of prior learning events include a second learning event that is not related to the first learning mode data and the plurality of prior learning events. The first learning mode data includes fourth learning mode data that is not related to the a priori learning event. For each fourth learning mode data, the fourth learning mode data is connected to the fourth learning mode data that best matches the fourth learning mode data. On the second learning event, the third correlation feature information is generated, including: If the online course education page is located in a learning mode sharing process, and based on the first correlation feature information and the second correlation feature information, it is determined that the plurality of prior learning events have data that is different from the first learning mode data. The related second learning event and the plurality of first learning mode data include fourth learning mode data that is not related to the a priori learning event. For each fourth learning mode data, the fourth learning mode data is associated with On the second learning event that the fourth learning mode data most matches, third correlation feature information is generated. 9. A digital education system, characterized in that it includes a processor and a computer-readable storage medium. The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are executed by the processor, the claims are realized. The big data push method based on digital education as described in any one of 1-8.