[go: up one dir, main page]

CN119205250A - Design of a new recommendation method for e-commerce platforms - Google Patents

Design of a new recommendation method for e-commerce platforms Download PDF

Info

Publication number
CN119205250A
CN119205250A CN202411209126.XA CN202411209126A CN119205250A CN 119205250 A CN119205250 A CN 119205250A CN 202411209126 A CN202411209126 A CN 202411209126A CN 119205250 A CN119205250 A CN 119205250A
Authority
CN
China
Prior art keywords
class
distance
data
user
commerce platform
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202411209126.XA
Other languages
Chinese (zh)
Inventor
李璋
李静
胥子悦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hubei University
Original Assignee
Hubei University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hubei University filed Critical Hubei University
Priority to CN202411209126.XA priority Critical patent/CN119205250A/en
Priority to NL2038659A priority patent/NL2038659B1/en
Publication of CN119205250A publication Critical patent/CN119205250A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0251—Targeted advertisements
    • G06Q30/0269—Targeted advertisements based on user profile or attribute
    • G06Q30/0271—Personalized advertisement
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0631—Recommending goods or services
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0475—Generative networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Economics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Game Theory and Decision Science (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a design of a novel recommendation method applied to an e-commerce platform, which is characterized in that influence factors are added in the calculation of distances to represent the relevance of multiple latitudes among objects, so that recommendation deviation caused by the fact that individual characteristics of users are not considered is avoided, basic preference information of users is acquired in a new user registration stage, the problem of cold start caused by the addition of the new users in the existing recommendation method is solved, the information quantity is added by filling coefficient matrixes of the users for object scoring by adopting an improved limited Boltzmann machine, and the recommendation deviation caused by incomplete metadata is solved, so that accuracy of recommendation results is effectively improved.

Description

Design of novel recommendation method applied to e-commerce platform
Technical Field
The invention relates to the technical field of electronic commerce, in particular to a design of a novel recommendation method applied to an electronic commerce platform.
Background
For the e-commerce platform, the recommendation method is core competitive. In the prior art, the recommendation method is mainly a content-based recommendation method, the content-based recommendation method extracts metadata of recommended articles according to user preference, finds out relevance among articles and rules thereof, and recommends similar articles to the user. However, the recommending method has the following defects that firstly, the recommended articles are modeled by the preference of the user, so that the quality of the recommended articles is excessively dependent on the completeness of the model, if the acquired metadata is incomplete, the difference between the recommended articles and the preference of the user is possibly large, secondly, the obtained articles are recommended only according to the labels of the articles, individual characteristics of the users are not considered, and therefore, the differences exist when the recommendation is carried out, thirdly, the addition of new users can generate a cold start problem, namely, the history behaviors are empty for the new users just entering the system, and therefore, the accurate recommendation cannot be carried out on the new users.
Disclosure of Invention
The invention aims to provide a design of a novel recommendation method applied to an e-commerce platform so as to solve the problems in the background technology.
The invention provides a novel recommendation method design applied to an electronic commerce platform, which comprises the following steps of analyzing the reasons of low accuracy of a basic recommendation method, designing an optimization scheme of the basic recommendation method, implementing the optimization scheme, adding the optimized recommendation scheme into the electronic commerce platform, testing software, and optimizing software;
In the first step, the reasons for low accuracy of the basic recommendation method based on the judgment similarity are analyzed in terms of data and classification;
In the second step, aiming at the data reasons put forward in the first step, the optimization scheme is specifically that basic preference of a new user is acquired in a registration stage of the new user, and in data modeling, an improved limited Boltzmann machine is used for filling a coefficient matrix of the user for scoring the object;
in the third step, according to the optimization scheme provided in the second step, the existing basic recommendation method is optimized;
in the fourth step, adding the recommendation scheme after the optimization treatment in the third step into the E-commerce platform;
in the fifth step, the recommended function of the e-commerce platform is tested, and the test method specifically comprises the steps of recruiting volunteers to simulate new user registration, and statistically analyzing satisfaction degree of the volunteers on a recommended result;
In the sixth step, the e-commerce platform software is optimized according to the test result in the fifth step, then the test in the fifth step is executed, and the test and optimization process is repeated until the satisfaction reaches a preset value.
Preferably, in the first step, the recommendation method based on the similarity specifically includes judging whether the article belongs to the preference of the user by judging the similarity of the article information of the merchant and the preference information of the user, if so, recommending the article to the user and recommending the user to the merchant corresponding to the article at the same time, and if not, entering the judgment of the next article until reaching the termination condition, wherein the termination condition specifically is that the preset recommended quantity is reached or the similarity judgment of all the articles is completed.
Preferably, in the first step, the reasons for the data aspect include that firstly, the metadata of the basic recommendation method is incomplete, so that the recommendation result lacks accuracy, and secondly, the new user information is missing, so that the recommendation result of the basic recommendation method lacks accuracy.
Preferably, in the first step, the reason for classification is that the basic recommendation method does not consider individual characteristic differences of users.
Preferably, in the second step, the specific method for obtaining the basic preference of the new user in the registration stage of the new user is that the new user is guided to complete filling of the basic information in the registration stage of the new user, thereby generating basic preference information by the information, then providing selection questions related to the preference for the user to select, and perfecting the basic preference information according to the user selection result.
Preferably, in the second step, the adding an influence factor in the calculation of the euclidean distance of the article is specifically:
Where i is constantly unequal to j, r ij is the pearson correlation coefficient for data samples i and j, max_r represents the maximum value of the pearson correlation coefficients for all data objects, and min_r represents the minimum value of the pearson correlation coefficients for all data objects;
Therefore, the distance after the influence factor is added is:
Dij＝dij×(1-gij+0.05)
d ij is a standard Euclidean distance, in order to prevent the distance obtained when the value of the influence factor is 1 from being 0, 0.05 is added at the back, after the calculation formula of the distance is obtained, the classification of the clusters can be started to be carried out on the whole data set, and after the completion, the isolated points are found out and eliminated.
Preferably, the classification of the class clusters is specifically that the distance between the smallest class and the class is required to be greater than or equal to the distance between the largest class inner element point and the center of the class, and the distance between the smallest class and the class is required to be greater than or equal to the distance between the smallest class and the center of the class:
the distance between the largest class interior element point and the class center:
Wherein D is the distance after the influence factors are added, c i and c j respectively represent the centers of two class clusters, and x i represents one data point belonging to the class cluster c i;
Average distance from class to class:
The minimum distance between classes is required to be smaller than or equal to the average value, and the maximum distance between the class inner element point and the center of the class is also smaller than the average value, so that coefficients lambda 1 and lambda 2 are necessarily present when the average distance between the classes is calculated by accumulation and then averaging
Setting a function:
H 1 in the function is the distance between the minimum class and class in the optimal state, h 2 is the distance between the maximum class inner element point and the class center in the optimal state, so the minimum value of the function y should be h 1 H 2 is 0, and in the clustering process, along with the continuous increase of the data volume, the divided classes can be approximately regarded as circles, the two classes are continuously close, the optimal condition is that the equal cutting and adding of the radii of the class circles of the two class clusters is equal to the distance between circle centers, and the following equation set is satisfied under the optimal condition:
The function y is carried into the cluster, and when lambda 1＝4/5,λ2 =2/5, y is the smallest, so that the data element can be divided from the cluster when the distance between the data element and the center of the cluster is larger than the average value of 4/5 times, and the data element is divided into the cluster when the distance between the data element and the center of the cluster is smaller than the average value of 2/5 times, and the data element is divided into k classes with obvious difference.
Preferably, the defining method of the isolated point is determined by density, the density is the number of neighbors of the data element in a certain radius range, in the process of continuous clustering, a clustering center is taken as a neighbor of the point, and thus, an average value of distances from the data point to the clustering center exists:
n is the total number of data sets employed, Representing the total set of 2 selected from the data set, k represents the number of cluster-like centers, and D is the distance after the influence factors are added;
if it meets the following conditions:
The data point i is the neighbor of the class j, all the points are circularly traversed, the proportion of all the data points in a certain radius range of the neighbor point of the data element is more than 25%, and the data element is an isolated point.
Preferably, in the second step, the improved limited boltzmann machine is used for filling a coefficient matrix of the user for scoring the object, specifically, the nodes of the visible layer and the hidden layer of the improved limited boltzmann machine are two-value variables, and the nodes are composed of 0 and 1, and the energy function of the model is as follows:
Wherein, the parameter theta of the improved limited Boltzmann machine consists of weight W between the visible layer and the hidden layer and offsets a and b corresponding to the visible layer and the hidden layer respectively, namely theta= (W, a, b), and n and m are the node numbers of the visible layer v and the hidden layer h respectively;
the joint distribution probability of two layers can be obtained:
Wherein z (θ) is a normalization factor:
From the above equation, after the model is input into the scoring matrix of the user, the activation probability of the units in the hidden layer is:
Where σ () is an activation function, set to:
the value of the visible layer i-th element can be obtained:
Preferably, in the step six, the optimizing method specifically optimizes the content of the selection questions to be filled in the new user registration stage, and increases the number of the selection questions so that the user preference information can be acquired more accurately.
Compared with the prior art, the method has the beneficial effects that the method has the advantages that the influence factors are added in the calculation of the distance to represent the multi-latitude correlation among the objects, so that recommendation deviation caused by the fact that individual characteristics of users are not considered is avoided, the method obtains basic preference information of the users in a new user registration stage, solves the problem of cold start caused by the addition of the new users in the existing recommendation method, and adds information quantity by filling coefficient matrixes of the users for object scoring through an improved limited Boltzmann machine, so that recommendation deviation caused by incomplete metadata is solved, and accuracy of recommendation results is effectively improved.
Drawings
FIG. 1 is a diagram showing the steps of a recommended method design according to the present invention;
FIG. 2 is a flowchart of a recommendation method of the present invention;
FIG. 3 is a flowchart illustrating a recommendation method according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-3, the design of a novel recommendation method applied to an e-commerce platform comprises the following steps of analyzing reasons for low accuracy of a basic recommendation method, designing an optimization scheme of the basic recommendation method, implementing the optimization scheme, adding the optimized recommendation scheme into the e-commerce platform, testing software, and optimizing software;
The method comprises the steps of judging whether an item belongs to the preference of a user or not by judging the similarity of item information of a merchant and preference information of the user, recommending the item to the user and recommending the user to the merchant corresponding to the item at the same time, judging the next item until a termination condition is reached, wherein the termination condition is that the preset recommended number is reached or the similarity judgment of all the items is completed;
In the second step, aiming at the data aspect reasons proposed in the first step, the optimization scheme is specifically that basic preference of a new user is acquired in a registration stage of the new user, in data modeling, an improved limited Boltzmann machine is used for filling a coefficient matrix of the user for scoring the object, aiming at the classification aspect reasons proposed in the first step, the optimization scheme is specifically that an influence factor is added in calculation of the Euclidean distance of the object to realize multi-dimensional consideration of individual difference, wherein the specific method for acquiring the basic preference of the new user in the registration stage of the new user is that the new user is firstly guided to complete filling of basic information in the registration stage of the new user, basic preference information is generated by the information, then selection questions related to the preference are provided for the user to select, the basic preference information is perfected according to a user selection result, and the influence factor is added in calculation of the Euclidean distance of the object:
Where i is constantly unequal to j, r ij is the pearson correlation coefficient for data samples i and j, max_r represents the maximum value of the pearson correlation coefficients for all data objects, and min_r represents the minimum value of the pearson correlation coefficients for all data objects;
Therefore, the distance after the influence factor is added is:
Dij＝dij×(1-gij+0.05)
d ij is a standard Euclidean distance, in order to prevent the distance obtained when the value of the influence factor is 1 from being 0, 0.05 is added at the back, after a calculation formula of the distance is obtained, the whole data set can be divided into clusters, isolated points are found out and eliminated after the completion of the classification, and the classification of the clusters is specifically that the distance between the minimum class and the class is required to be greater than or equal to the distance between the maximum class inner element point and the center of the class, and the distance between the minimum class and the class is required to be greater than or equal to the distance between the minimum class and the center of the class:
the distance between the largest class interior element point and the class center:
Wherein D is the distance after the influence factors are added, c i and c j respectively represent the centers of two class clusters, and x i represents one data point belonging to the class cluster c i;
Average distance from class to class:
The minimum distance between classes is required to be smaller than or equal to the average value, and the maximum distance between the class inner element point and the center of the class is also smaller than the average value, so that coefficients lambda 1 and lambda 2 are necessarily present when the average distance between the classes is calculated by accumulation and then averaging
Setting a function:
H 1 in the function is the distance between the minimum class and class in the optimal state, h 2 is the distance between the maximum class inner element point and the class center in the optimal state, so the minimum value of the function y should be h 1 H 2 is 0, and in the clustering process, along with the continuous increase of the data volume, the divided classes can be approximately regarded as circles, the two classes are continuously close, the optimal condition is that the equal cutting and adding of the radii of the class circles of the two class clusters is equal to the distance between circle centers, and the following equation set is satisfied under the optimal condition:
The method is characterized in that when lambda 1＝4/5,λ2 =2/5, y is the smallest, so that a data element is divided from the class when the distance between the data element and the center of the class cluster is larger than the average value of 4/5 times, when the distance between the data element and the center of the class cluster is smaller than the average value of 2/5 times, the data element is classified into the class, a complete data set is divided into k classes with obvious difference, the definition method of isolated points is judged by density, the density is the number of neighbors of the data element in a certain radius range, and in the process of continuous clustering, the neighbors with the cluster center as the points are taken, so that the average value of the distances from the data point to the cluster center exists:
n is the total number of data sets employed, Representing the total set of 2 selected from the data set, k represents the number of cluster-like centers, and D is the distance after the influence factors are added;
if it meets the following conditions:
The method is characterized in that a data point i is a neighbor of a class j, all points are circularly traversed, the ratio of the neighbor points of the data element to all data points in a certain radius range is more than 25%, the data element is an isolated point, a coefficient matrix (shown in a table 1) for scoring an article by a user is filled by using an improved limited Boltzmann machine, specifically, nodes of a visible layer and a hidden layer of the improved limited Boltzmann machine are binary variables, the nodes consist of 0 and 1, and the energy function of the model is as follows:
Wherein, the parameter theta of the improved limited Boltzmann machine consists of weight W between the visible layer and the hidden layer and offsets a and b corresponding to the visible layer and the hidden layer respectively, namely theta= (W, a, b), and n and m are the node numbers of the visible layer v and the hidden layer h respectively;
the joint distribution probability of two layers can be obtained:
Wherein z (θ) is a normalization factor:
From the above equation, after the model is input into the scoring matrix of the user, the activation probability of the units in the hidden layer is:
Where σ () is an activation function, set to:
the value of the visible layer i-th element can be obtained:
in the third step, according to the optimization scheme provided in the second step, the existing basic recommendation method is optimized;
in the fourth step, adding the recommendation scheme after the optimization treatment in the third step into the E-commerce platform;
in the fifth step, the recommended function of the e-commerce platform is tested, and the test method specifically comprises the steps of recruiting volunteers to simulate new user registration, and statistically analyzing satisfaction degree of the volunteers on a recommended result;
In the sixth step, the e-commerce platform software is optimized according to the test result in the fifth step, then the test in the fifth step is executed, and the test and optimization process is repeated until the satisfaction reaches a preset value, wherein the optimization method specifically comprises the steps of optimizing the content of the selection questions to be filled in the new user registration stage, and increasing the number of the selection questions so that the more accurate user preference information can be obtained.
Table 1 coefficient matrix for user scoring of items
Note that I represents items, u represents users, and r represents the unused user's score for different items.
Based on the above, the method has the advantages that when the method is used, firstly, influence factors are added in the calculation of the distance to represent the relevance of multiple latitudes among objects, recommendation deviation caused by individual characteristics of users is avoided, secondly, the problem of cold start is solved by acquiring basic preference information of the users, furthermore, the information quantity is added by filling coefficient matrixes of the scores of the users for the objects through an improved limited Boltzmann machine, so that recommendation deviation caused by incomplete metadata is solved, and interference factors existing in the method can be reduced by removing isolated points after the distance classification.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.

Claims (10)

1. The design of the novel recommendation method applied to the e-commerce platform comprises the following steps of analyzing the reason of low accuracy of a basic recommendation method, designing an optimization scheme of the basic recommendation method, implementing the optimization scheme, adding the optimized recommendation scheme into the e-commerce platform, testing software, optimizing the software, and the design method is characterized in that:
In the first step, the reasons for low accuracy of the basic recommendation method based on the judgment similarity are analyzed in terms of data and classification;
In the second step, aiming at the data reasons put forward in the first step, the optimization scheme is specifically that basic preference of a new user is acquired in a registration stage of the new user, and in data modeling, an improved limited Boltzmann machine is used for filling a coefficient matrix of the user for scoring the object;
in the third step, according to the optimization scheme provided in the second step, the existing basic recommendation method is optimized;
in the fourth step, adding the recommendation scheme after the optimization treatment in the third step into the E-commerce platform;
in the fifth step, the recommended function of the e-commerce platform is tested, and the test method specifically comprises the steps of recruiting volunteers to simulate new user registration, and statistically analyzing satisfaction degree of the volunteers on a recommended result;
In the sixth step, the e-commerce platform software is optimized according to the test result in the fifth step, then the test in the fifth step is executed, and the test and optimization process is repeated until the satisfaction reaches a preset value.
2. The method for recommending the electronic commerce platform according to claim 1 is characterized in that in the first step, the recommending method based on the similarity is characterized in that whether the article belongs to the preference of the user is judged by judging the similarity of the article information of the merchant and the preference information of the user, if so, the article is recommended to the user, and meanwhile, the user is recommended to the merchant corresponding to the article, and if not, the judgment of the next article is entered until a termination condition is reached, wherein the termination condition is that the preset recommended number is reached or the similarity judgment of all the articles is completed.
3. The method for designing the novel recommendation method for the e-commerce platform according to claim 1, wherein in the first step, the data reasons comprise that firstly, the recommendation result lacks accuracy due to incomplete metadata of the basic recommendation method, and secondly, the recommendation result lacks accuracy due to new user information deficiency.
4. The method for designing the novel recommendation method for the e-commerce platform according to claim 1, wherein in the first step, the classification is performed for the reason that the basic recommendation method does not consider individual characteristic differences of users.
5. The method for designing the novel recommendation method for the e-commerce platform according to claim 1, wherein in the second step, the specific method for acquiring the basic preference of the new user in the registration stage of the new user is that the new user is guided to complete filling of basic information in the registration stage of the new user, so that basic preference information is generated by the information, then selection questions related to the preference are provided for the user to select, and the basic preference information is perfected according to the user selection result.
6. The method for designing the novel recommendation method for the e-commerce platform according to claim 1, wherein in the second step, an influence factor is added to the calculation of the Euclidean distance of the object, specifically:
Where i is constantly unequal to j, r ij is the pearson correlation coefficient for data samples i and j, max_r represents the maximum value of the pearson correlation coefficients for all data objects, and min_r represents the minimum value of the pearson correlation coefficients for all data objects;
Therefore, the distance after the influence factor is added is:
Dij＝dij×(1-gij+0.05)
d ij is a standard Euclidean distance, in order to prevent the distance obtained when the value of the influence factor is 1 from being 0, 0.05 is added at the back, after the calculation formula of the distance is obtained, the classification of the clusters can be started to be carried out on the whole data set, and after the completion, the isolated points are found out and eliminated.
7. The design of the novel recommendation method applied to the e-commerce platform according to claim 6, wherein the classification of the class clusters is specifically that the distance between the smallest class and the class is greater than or equal to the distance between the largest class inner element point and the center of the class, and the distance between the smallest class and the class is:
the distance between the largest class interior element point and the class center:
Wherein D is the distance after the influence factors are added, c i and c j respectively represent the centers of two class clusters, and x i represents one data point belonging to the class cluster c i;
Average distance from class to class:
The minimum distance between classes is required to be smaller than or equal to the average value, and the maximum distance between the class inner element point and the center of the class is also smaller than the average value, so that coefficients lambda 1 and lambda 2 are necessarily present when the average distance between the classes is calculated by accumulation and then averaging
Setting a function:
H 1 in the function is the distance between the minimum class and class in the optimal state, h 2 is the distance between the maximum class inner element point and the class center in the optimal state, so the minimum value of the function y should be h 1 H 2 is 0, and in the clustering process, along with the continuous increase of the data volume, the divided classes can be approximately regarded as circles, the two classes are continuously close, the optimal condition is that the equal cutting and adding of the radii of the class circles of the two class clusters is equal to the distance between circle centers, and the following equation set is satisfied under the optimal condition:
The function y is carried into the cluster, and when lambda 1＝4/5,λ2 =2/5, y is the smallest, so that the data element can be divided from the cluster when the distance between the data element and the center of the cluster is larger than the average value of 4/5 times, and the data element is divided into the cluster when the distance between the data element and the center of the cluster is smaller than the average value of 2/5 times, and the data element is divided into k classes with obvious difference.
8. The method for designing a novel recommendation method for an e-commerce platform according to claim 6, wherein the defining method of isolated points is determined by density, the density is calculated by calculating the number of neighbors of the data element in a certain radius range, and in the process of continuous clustering, a clustering center is taken as a neighbor of a point, so that an average value of distances from a data point to the clustering center exists:
n is the total number of data sets employed, Representing the total set of 2 selected from the data set, k represents the number of cluster-like centers, and D is the distance after the influence factors are added;
if it meets the following conditions:
The data point i is the neighbor of the class j, all the points are circularly traversed, the proportion of all the data points in a certain radius range of the neighbor point of the data element is more than 25%, and the data element is an isolated point.
9. The method according to claim 1, wherein in the second step, the improved Boltzmann machine is used for filling a coefficient matrix of the scoring of the object by the user, wherein nodes of a visible layer and a hidden layer of the improved Boltzmann machine are binary variables, and the nodes consist of 0 and 1, and the energy function of the model is as follows:
Wherein, the parameter theta of the improved limited Boltzmann machine consists of weight W between the visible layer and the hidden layer and offsets a and b corresponding to the visible layer and the hidden layer respectively, namely theta= (W, a, b), and n and m are the node numbers of the visible layer v and the hidden layer h respectively;
the joint distribution probability of two layers can be obtained:
Wherein z (θ) is a normalization factor:
From the above equation, after the model is input into the scoring matrix of the user, the activation probability of the units in the hidden layer is:
Where σ () is an activation function, set to:
the value of the visible layer i-th element can be obtained:
10. The method for designing the novel recommendation method for the e-commerce platform according to claim 1, wherein in the step six, the optimization method is specifically that the content of the selection questions to be filled in a new user registration stage is optimized, and the number of the selection questions is increased, so that more accurate user preference information can be obtained.
CN202411209126.XA 2024-08-30 2024-08-30 Design of a new recommendation method for e-commerce platforms Pending CN119205250A (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202411209126.XA CN119205250A (en) 2024-08-30 2024-08-30 Design of a new recommendation method for e-commerce platforms
NL2038659A NL2038659B1 (en) 2024-08-30 2024-09-18 Design of a novel recommendation method applied to e-commerce platforms

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202411209126.XA CN119205250A (en) 2024-08-30 2024-08-30 Design of a new recommendation method for e-commerce platforms

Publications (1)

Publication Number Publication Date
CN119205250A true CN119205250A (en) 2024-12-27

Family

ID=93155053

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202411209126.XA Pending CN119205250A (en) 2024-08-30 2024-08-30 Design of a new recommendation method for e-commerce platforms

Country Status (2)

Country Link
CN (1) CN119205250A (en)
NL (1) NL2038659B1 (en)

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108475392B (en) * 2015-11-24 2022-12-06 波特馥公司 System and method for tracking consumer taste preferences

Also Published As

Publication number Publication date
NL2038659A (en) 2024-10-24
NL2038659B1 (en) 2026-01-12

Similar Documents

Publication Publication Date Title
CN103377250B (en) Top k based on neighborhood recommend method
CN109902823B (en) Model training method and device based on generation countermeasure network
CN117112784B (en) A method for realizing document identification and technology path evolution
CN119621891B (en) Method, system and device for enhancing factuality of legal question answering based on large language model
CN108509588B (en) Lawyer evaluation method and recommendation method based on big data
CN117951280A (en) A generative large model-based intelligent question answering learning method
CN118246990A (en) Virtual twin product aesthetics and cultural style evaluation system and method based on big data
CN116049543A (en) Hybrid recommendation method, system and storage medium for comprehensive energy efficiency service business
CN117010373B (en) Recommendation method for category and group to which asset management data of power equipment belong
CN112598405B (en) Business project data management method and system based on big data
CN120277269A (en) Social-integrated intelligent body and recommendation simulation environment construction method
CN119205250A (en) Design of a new recommendation method for e-commerce platforms
CN120298079A (en) A product recommendation method based on personality traits and FCM clustering optimization
CN112070336A (en) Manufacturing industry information quantitative analysis method and device based on analytic hierarchy process
KR20240076323A (en) Reverse direct purchase sales products by country and method of recommending sales price
CN115358323A (en) A feature screening method and device
CN116188027A (en) Commodity semantic vector generation method and device and commodity recall method and system
CN120218979B (en) Product strategy lifting method based on LSGDM-SHAP
CN114595151B (en) Prototype Software User Experience Evaluation Method
CN118657593B (en) Knowledge graph and price analysis-based intelligent material recommendation method and system
CN118036602B (en) False comment identification method and device
CN111651512A (en) Multisource heterogeneous commodity feature weight solving method and device based on semi-supervised learning
CN119358519B (en) An evaluation report generation system based on artificial intelligence and big data analysis
CN120047219B (en) Community Internet Mall Information Recommendation Method Based on User Behavior Big Data
CN118820813B (en) Product cluster analysis method based on deep learning model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication