CN113989075B - Methods for predicting future technological knowledge flows - Google Patents
Methods for predicting future technological knowledge flows Download PDFInfo
- Publication number
- CN113989075B CN113989075B CN202111186872.8A CN202111186872A CN113989075B CN 113989075 B CN113989075 B CN 113989075B CN 202111186872 A CN202111186872 A CN 202111186872A CN 113989075 B CN113989075 B CN 113989075B
- Authority
- CN
- China
- Prior art keywords
- technology
- diffusion
- technical
- time period
- absorption
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000005516 engineering process Methods 0.000 claims abstract description 144
- 238000009792 diffusion process Methods 0.000 claims abstract description 87
- 239000013598 vector Substances 0.000 claims abstract description 65
- 230000004931 aggregating effect Effects 0.000 claims abstract description 12
- 238000010521 absorption reaction Methods 0.000 claims description 70
- 230000006870 function Effects 0.000 claims description 19
- 230000009466 transformation Effects 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 14
- 230000003993 interaction Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000012546 transfer Methods 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 4
- 230000002457 bidirectional effect Effects 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 230000006403 short-term memory Effects 0.000 claims 1
- 238000011161 development Methods 0.000 abstract description 4
- 230000002452 interceptive effect Effects 0.000 abstract description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000000593 degrading effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000006116 polymerization reaction Methods 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/18—Legal services
- G06Q50/184—Intellectual property management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/38—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/382—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using citations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Tourism & Hospitality (AREA)
- Technology Law (AREA)
- Evolutionary Computation (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Databases & Information Systems (AREA)
- Library & Information Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明公开了一种预测未来技术知识流动的方法，包括：步骤1、获取每一个时间段上的授权的专利及其所属的技术分类，以及每个专利的引用专利文献；步骤2、将所述专利间的引用聚合成技术间的流动，构建每一年的技术知识流动TKF网络；步骤3、将所述专利数据通过计算得到每一个时间段内技术领域在扩散能力和吸收能力上的初始特征向量；步骤4、将上述每一个时间段上的技术的初始特征向量输入技术流动预测框架，得到下一个时间段上的技术扩散能力的特征向量与吸收能力的特征向量，并将不同技术的扩散能力与吸收能力进行匹配，得到最终的技术之间的流动趋势。该方法能够高效且准确地实现对未来技术领域之间的交互关系以及发展趋势进行预测。
The present invention discloses a method for predicting the future flow of technological knowledge, including: step 1, obtaining the authorized patents and their technical classifications in each time period, as well as the cited patent documents of each patent; step 2, aggregating the citations between the patents into the flow between technologies, and constructing the technical knowledge flow TKF network of each year; step 3, calculating the patent data to obtain the initial feature vectors of the diffusion ability and absorptive ability of the technical field in each time period; step 4, inputting the initial feature vectors of the technology in each time period into the technology flow prediction framework, obtaining the feature vectors of the technology diffusion ability and the feature vectors of the absorptive ability in the next time period, and matching the diffusion ability and absorptive ability of different technologies to obtain the final flow trend between technologies. The method can efficiently and accurately predict the interactive relationship and development trend between future technical fields.
Description
Technical Field
The invention relates to the technical field of big data analysis, in particular to a method for predicting future technical knowledge flow.
Background
Technical knowledge flow (Technological Knowledge Flow, TKF for short) refers to the directional flow of knowledge from one technical field to another, reflecting one interaction between two different technical fields. In the trend of economic globalization, many high and new technical enterprises face severe innovation pressure, and in order to maintain competitive advantage, they must grasp the technical development trend, and are in progress. In this respect, an effective method is to predict the flow between different technical fields in the future, thereby grasping the development trend of the technology and establishing effective research and development strategy.
TKF analysis is often performed in large-scale scientific literature, where patent mining plays an important role. In the patent arts, each patent often references an already issued patent and at the time of each patent application is assigned one or more technical classification codes according to the patent classification system. Taking the most commonly used joint patent classification system (CPC) at present as an example, each classification code can be regarded as a technical field, and different technologies can be connected in two ways, one is a hierarchical structure inherent to CPC classification, and the other is connected by patent citation. Thus, with the CPC taxonomy and patent citation, we can construct a annual technical knowledge flow graph and predict future technical knowledge flow trends.
Disclosure of Invention
The invention aims to provide a method for predicting future technical knowledge flow, which can efficiently and accurately predict the interactive relation and development trend among the future technical fields.
To achieve the above object, the present invention provides a method of predicting a future technical knowledge flow, the method comprising:
Step 1, acquiring the issued patents and technical classifications to which the issued patents belong in each time period and the cited patent literature of each patent;
step 2, aggregating the references among the patents into flows among technologies, and constructing a technical knowledge flow TKF network of each year;
Step 3, calculating the patent data to obtain initial feature vectors of the technical field in the diffusion capacity and the absorption capacity in each time period;
and 4, inputting the initial feature vectors of the technologies in each time period into a technology flow prediction framework to obtain feature vectors of the technology diffusion capacity and the absorption capacity in the next time period, and matching the diffusion capacity and the absorption capacity of different technologies to obtain the flow trend among the final technologies.
Preferably, the technique diffusivity initial feature vector d in step 3 includes:
The technology growth rate, namely the ratio of the authorized quantity of a technology to the total authorized quantity of the technology in the last years;
the amount of technology spread, i.e., the amount of other technical fields to which the knowledge contained in a technology is spread;
The technology diffusivity, i.e., the proportion of the diffusion of a technology to the total diffusion of all technologies.
Preferably, the initial feature vector a of the technical absorption capacity in step3 includes:
The technology growth rate, namely the ratio of the authorized quantity of a technology to the total authorized quantity of the technology in the last years;
the technical absorption, i.e. the amount of knowledge that a technology absorbs comes from other technical fields;
Technical absorptivity, i.e. the proportion of the absorption of a technology to the total absorption of all technologies.
Preferably, step 4 comprises:
Inputting an initial feature vector d of the technical diffusion capability and an initial feature vector a of the technical absorption capability of the technology in each time period into a high-order interaction module HOI of a technical flow prediction framework, and calculating to obtain a feature vector d o of the diffusion capability and a feature vector a o of the absorption capability after surrounding high-order neighbors are polymerized in the technical field;
Inputting the feature vectors d o and a o into a hierarchical transfer module, and obtaining technical diffusion capacity and absorption capacity feature vectors containing hierarchical structure information specific to a technical system through calculation;
Inputting the characteristic vectors of the technical diffusion capacity and the absorption capacity of each time period into a technical flow tracking module to obtain characteristic vectors of the diffusion capacity and the absorption capacity of the technical field in the next time period;
Matching the diffusion capacity and the absorption capacity of the technology in the future time period to obtain the flow probability between two different technical fields;
and inputting the obtained probability and the true value into a loss function, and carrying out gradient update.
Preferably, the calculation formula of the high-order interaction module in step 4 is:
Wherein, AndThe representation technique U i embeds the diffuse and absorptive representations of the k-th layer neighbors,AndRepresenting the outgoing and incoming neighbor sets of technique U i, respectively, after aggregating the higher-order neighbor features of the K-layer, obtaining an updated technique diffusivity representationAnd absorbency representation
Wherein, alpha k is equal to or greater than 0, which represents the importance of the k-th layer neighbor, namely the influence of the k-th layer neighbor on the technology U i.
Preferably, the hierarchical delivery module in step 4 includes:
The upward convergence module is used for converging the information of the lower nodes to the upper nodes, and the module adopts an attention mechanism to automatically learn the weights of the lower nodes to represent the importance of different lower nodes, and the formula is as follows:
Where w ij denotes the weight of the lower node j relative to the upper node i, Is a transformation matrix which is a function of the transformation matrix,Is the offset vector, |g i | is the number of all lower-level technology nodes contained in the upper-level technology node g i, | ij is the normalized weight,Is the final representation of the upper node i;
a downward updating module for updating the upper node Is updated to the lower nodeThe calculation formula is as follows:
Wherein, Is an updated representation of the lower level node,Is a transformation matrix which is a function of the transformation matrix,Is the offset vector and σ is the activation function.
Preferably, in the technical flow tracking module of step 4, a bidirectional long and short time memory model BiLSTM is used for modeling the dynamic characteristics, and the specific formula is as follows:
The diffusion characteristic d= [ d 1,d2,...,dT ] over the continuous period T is input BiLSTM,
dt★＝ht
Obtaining a diffusion profile over a period of T+1
According to the technical scheme, initial diffusion characteristics and absorption characteristics of each technical field in a continuous time period are firstly extracted, wherein the initial diffusion characteristics and absorption characteristics comprise the growth rate of the technology, the diffusion capacity and absorption capacity of the technology and the diffusion rate and absorption rate of the technology, then the initial diffusion characteristics and the absorption characteristics are input into a preset technology flow frame, the characteristics sequentially pass through a high-order interaction module and a hierarchical transmission module of the technology to obtain characteristic vectors of the technology in diffusion and absorption aspects in each time period, then the diffusion capacity vectors and absorption capacity vectors of the technology in each time period are input into a technology flow tracking module to obtain diffusion capacity and absorption capacity vectors of the technology in future time periods, and finally the flow probability between two different technical fields is obtained by matching the diffusion capacity and the absorption capacity of the technology. The hierarchical structure property of the existing patent classification system and the high-order flow among technologies are fully utilized, a technology flow network based on patent citation is constructed by using a method of a graph neural network, and the future technology flow direction is predicted according to the dynamic characteristics of the technology. The method overcomes the defects that the prior method mainly depends on expert experience and needs to consume a large amount of manpower, and the characteristics of technical flow and the characteristics of patent hierarchy structure cannot be fully utilized, and can efficiently and accurately predict the future technical flow direction.
Additional features and advantages of the invention will be set forth in the detailed description which follows.
Drawings
The accompanying drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate the invention and together with the description serve to explain, without limitation, the invention. In the drawings:
FIG. 1 is a flow chart of a method of predicting future technical knowledge flow in accordance with the present invention;
fig. 2 is a schematic diagram of a method for predicting future technical knowledge flow according to the present invention.
Detailed Description
The following describes specific embodiments of the present invention in detail with reference to the drawings. It should be understood that the detailed description and specific examples, while indicating and illustrating the invention, are not intended to limit the invention.
Referring to fig. 2, the present invention provides a method of predicting future technical knowledge flow, comprising:
Step 1, acquiring the issued patents and technical classifications to which the issued patents belong in each time period and the cited patent literature of each patent;
step 2, aggregating the references among the patents into flows among technologies, and constructing a technical knowledge flow TKF network of each year;
Step 3, calculating the patent data to obtain initial feature vectors of the technical field in the diffusion capacity and the absorption capacity in each time period;
and 4, inputting the initial feature vectors of the technologies in each time period into a technology flow prediction framework to obtain feature vectors of the technology diffusion capacity and the absorption capacity in the next time period, and matching the diffusion capacity and the absorption capacity of different technologies to obtain the flow trend among the final technologies.
Wherein, the technique diffusivity initial feature vector d in step 3 includes:
The technology growth rate, namely the ratio of the authorized quantity of a technology to the total authorized quantity of the technology in the last years;
the amount of technology spread, i.e., the amount of other technical fields to which the knowledge contained in a technology is spread;
The technology diffusivity, i.e., the proportion of the diffusion of a technology to the total diffusion of all technologies.
The initial feature vector a of the technical absorption capacity in step 3 includes:
The technology growth rate, namely the ratio of the authorized quantity of a technology to the total authorized quantity of the technology in the last years;
the technical absorption, i.e. the amount of knowledge that a technology absorbs comes from other technical fields;
Technical absorptivity, i.e. the proportion of the absorption of a technology to the total absorption of all technologies.
Specifically, step 4 includes:
Inputting an initial feature vector d of the technical diffusion capability and an initial feature vector a of the technical absorption capability of the technology in each time period into a high-order interaction module HOI of a technical flow prediction framework, and calculating to obtain a feature vector d o of the diffusion capability and a feature vector a o of the absorption capability after surrounding high-order neighbors are polymerized in the technical field;
Inputting the feature vectors d o and a o into a hierarchical transfer module, and obtaining technical diffusion capacity and absorption capacity feature vectors containing hierarchical structure information specific to a technical system through calculation;
Inputting the characteristic vectors of the technical diffusion capacity and the absorption capacity of each time period into a technical flow tracking module to obtain characteristic vectors of the diffusion capacity and the absorption capacity of the technical field in the next time period;
Matching the diffusion capacity and the absorption capacity of the technology in the future time period to obtain the flow probability between two different technical fields;
and inputting the obtained probability and the true value into a loss function, and carrying out gradient update.
The calculation formula of the high-order interaction module in the step 4 is as follows:
Wherein, AndThe representation technique U i embeds the diffuse and absorptive representations of the k-th layer neighbors,AndRepresenting the outgoing and incoming neighbor sets of technique U i, respectively, after aggregating the higher-order neighbor features of the K-layer, obtaining an updated technique diffusivity representationAnd absorbency representation
Wherein, alpha k is equal to or greater than 0, which represents the importance of the k-th layer neighbor, namely the influence of the k-th layer neighbor on the technology U i.
The hierarchical delivery module in step 4 includes:
The upward convergence module is used for converging the information of the lower nodes to the upper nodes, and the module adopts an attention mechanism to automatically learn the weights of the lower nodes to represent the importance of different lower nodes, and the formula is as follows:
Where w ij denotes the weight of the lower node j relative to the upper node i, Is a transformation matrix which is a function of the transformation matrix,Is the offset vector, |g i | is the number of all lower-level technology nodes contained in the upper-level technology node g i, | ij is the normalized weight,Is the final representation of the upper node i;
a downward updating module for updating the upper node Is updated to the lower nodeThe calculation formula is as follows:
Wherein, Is an updated representation of the lower level node,Is a transformation matrix which is a function of the transformation matrix,Is the offset vector and σ is the activation function.
In addition, in the technical flow tracking module in step 4, a bidirectional long-short-time memory model BiLSTM is used for modeling of dynamic characteristics, and a specific formula is as follows:
The diffusion characteristic d= [ d 1,d2,...,dT ] over the continuous period T is input BiLSTM,
dt★＝ht
Obtaining a diffusion profile over a period of T+1
Specifically, as shown in fig. 1, in one embodiment of the present invention, the following steps are sequentially performed:
S101, acquiring the issued patents and technical classifications to which the issued patents belong in each time period, and the cited patent literature of each patent.
In the method provided by the embodiment of the invention, according to the patent data provided by USPTO of the U.S. patent and trademark office and the technical classification provided by CPC of the united patent classification system, the new patents and the technical classification of the new patents and the technical classification and the cited other patents are extracted every year according to years.
S102, aggregating the references among the patents into inter-technology flow, and constructing a technology knowledge flow TKF network of each year.
In the method provided by the embodiment of the invention, if the patent contained in the technology A refers to the patent contained in the technology B for n times, it is considered that there may be a knowledge flow from the technology B to the technology A, wherein n represents the intensity of the flow. Clearly, the larger n, the more knowledge flows between technology a and technology B. In order to remove noise effects and limit the size of the technical flow network, N is thresholded such that the corresponding knowledge flow can only act as a side of the technical knowledge flow network if N > N. The value of N is set according to the specific requirement in the implementation process, and the magnitude of the specific value is determined according to the knowledge flow strength requirement between the technology a and the technology B.
And S103, calculating the patent data to obtain initial characteristic vectors of the technical field in the diffusion capacity and the absorption capacity in each time period.
In the method provided by the embodiment of the invention, the initial eigenvectors of the diffusion capacity and the absorption capacity respectively comprise three values, namely a technical growth rate, a technical diffusion quantity, a technical diffusion rate, a technical growth rate, a technical absorption quantity and a technical absorption rate. Order theAndRepresenting the diffusion and absorption characteristics, respectively, of the technique U i over time t, willAndAs input to the subsequent module.
S104, inputting the initial feature vector d and a of the technology in each time period into a high-order interaction module HOI of a technology flow prediction framework, and calculating to obtain a diffusion capability feature vector d o and an absorption capability feature vector a o after surrounding high-order neighbors are aggregated in the technical field.
In the method provided by the embodiment of the invention, the edges in the TKF network graph can be expressed as the interaction of the source node diffusion capability and the target node absorption capability, so that the two characteristics of the technology are updated by adopting the idea of the interactive graph neural network. In particular, the diffusion characteristics of a technical field are updated by aggregating the absorption characteristics of adjacent technical fields, and vice versa. A simple weighted sum operation is chosen here as operator, taking into account efficiency and effect. The polymerization process is as follows:
Wherein, AndThe representation technique U i embeds the diffuse and absorptive representations of the k-th layer neighbors,AndRepresenting the outgoing and incoming neighbor sets of technology U i, respectively. After aggregating the higher-order neighbor features of the K-layer, an updated technology diffusivity representation is obtainedAnd absorbency representation
Wherein alpha k is equal to or greater than 0, which represents the importance of the k-th layer neighbor, namely the influence of the second rolling neighbor on the technology U i.
S105, inputting the feature vectors into a hierarchical transfer module, and obtaining the feature vectors of the technical diffusion capability and the absorption capability containing the hierarchical structure information specific to the technical system through calculation.
In the method provided by the embodiment of the invention, the technology can be connected through the hierarchical structure of the CPC system. Thus, a more comprehensive technical representation can be generated by using the hierarchical relationship between technologies, and more accurate TKF prediction is realized. Since the above category has an inclusive relationship with the below category, the above category can be represented by aggregating the information of the below category. On the other hand, the information of the higher category can also supplement incomplete information in the lower category, and a hierarchical delivery module is constructed for the purpose of describing the interaction between the upper category and the lower category. The hierarchical transfer module consists of two parts, namely an upward transfer module and a downward update module. In this embodiment, each Section of the CPC may form a dataset, as shown in fig. 2, focusing mainly on the hierarchical structure between the three levels, subgroup, groups and subsection, respectively.
The upward passing module aims to aggregate information from a lower level to a higher level, i.e., subgroup to group, group to subsection. Here, taking the upward pass from subgroup to group as an example, the diffusion characteristics for a group level node are represented by aggregating its diffusion characteristics for its children at the subgroup level. To distinguish the importance of different child nodes, an attention mechanism is applied to automatically learn the weights of the child nodes. The weights representing node j of the subgroup hierarchy to node i of the group hierarchy, denoted by w ij, can be calculated as:
Wherein, Is a transformation matrix which is a function of the transformation matrix,Is the offset vector, |g i | is the number of all lower-level technology nodes contained in the upper-level technology node g i, | ij is the normalized weight,Is an updated diffusion characteristic representation of group level node i. Likewise, the absorption features of technique i at the group level can be obtainedThen, through the upward transfer process from the group level to the subsection level, the diffusion characterization of the technology i at the subsection level is obtainedAnd absorption characterization
In the downward update phase, the lower nodes, i.e., subsection to group and group to subgroup, need to be updated with the information of the upper nodes. Here, taking the subsection level to group level update process as an example, to update the diffusion representation of the nodes in the group level, the diffusion feature of the current node and the diffusion feature of its parent node in the subsection level are spliced, and then nonlinear transformation is adopted to obtain the enhanced diffusion representation:
Wherein, Is a node representation of the updated group hierarchy,Is a transformation matrix which is a function of the transformation matrix,Is the offset vector and σ is the activation function. Likewise, through the information updating from the group level to the subgroup level, the diffusion characteristics of the technology j updated after passing through the level transmission module on the subgroup level can be obtainedAnd absorption features
S106, inputting the characteristic vectors of the technical diffusion capacity and the absorption capacity of each time period into a technical flow tracking module to obtain the characteristic vectors of the diffusion capacity and the absorption capacity of the technical field in the next time period.
In the embodiment of the application, a bidirectional long-short-time memory model BiLSTM is used for modeling dynamic characteristics, and a diffusion characteristic d= [ d 1,d2,...,dT ] in a continuous time period T is input BiLSTM, wherein the specific formula is as follows:
dt★＝ht
Obtaining a diffusion profile over a period of T+1
And S107, matching the diffusion capacity and the absorption capacity of the technology in the future time period to obtain the flow probability between two different technical fields.
In this embodiment, the diffusion characteristic at time T+1 is setAnd absorption featuresAnd performing inner product operation, wherein the formula is as follows: The probability of technology i flowing to technology j can be finally obtained for the t+1 time period.
S108, inputting the obtained probability and the true value into a loss function, and carrying out gradient update.
In this embodiment, because the completely new technical flows are difficult to predict, they have never been presented before, and therefore a well-designed objective function is still needed to train the model. In order to more accurately predict new edges without degrading accuracy, a loss function is designed for edges that occur for each time window, with the following formula:
Wherein, Representing a loss function that predicts entirely new edges,Representing a loss function that predicts all edges that occur,Is the total loss function, alpha representsAndAt the position ofThe specific gravity of the mixture is calculated.
The method comprises the steps of firstly extracting initial diffusion characteristics and absorption characteristics of each technical field in a continuous time period, wherein the initial diffusion characteristics and absorption characteristics comprise the growth rate of the technology, the diffusion capacity and absorption capacity of the technology and the diffusion capacity and absorption capacity of the technology, then inputting the initial diffusion characteristics and absorption characteristics into a preset technology flow frame, sequentially passing through a high-order interaction module and a hierarchical transfer module of the technology to obtain characteristic vectors of the technology in diffusion and absorption aspects in each time period, then inputting the diffusion capacity vectors and absorption capacity vectors of the technology in each time period into a technology flow tracking module to obtain diffusion capacity and absorption capacity vectors of the technology in future time periods, and finally obtaining flow probability between two different technical fields by matching the diffusion capacity and the absorption capacity of the technology. The hierarchical structure property of the existing patent classification system and the high-order flow among technologies are fully utilized, a technology flow network based on patent citation is constructed by using a method of a graph neural network, and the future technology flow direction is predicted according to the dynamic characteristics of the technology. The method overcomes the defects that the prior method mainly depends on expert experience and needs to consume a large amount of manpower, and the characteristics of technical flow and the characteristics of patent hierarchy structure cannot be fully utilized, and can efficiently and accurately predict the future technical flow direction.
The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the specific details of the above embodiments, and various simple modifications can be made to the technical solution of the present invention within the scope of the technical concept of the present invention, and all the simple modifications belong to the protection scope of the present invention.
In addition, the specific features described in the above embodiments may be combined in any suitable manner, and in order to avoid unnecessary repetition, various possible combinations are not described further.
Moreover, any combination of the various embodiments of the invention can be made without departing from the spirit of the invention, which should also be considered as disclosed herein.
Claims (5)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111186872.8A CN113989075B (en) | 2021-10-12 | 2021-10-12 | Methods for predicting future technological knowledge flows |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111186872.8A CN113989075B (en) | 2021-10-12 | 2021-10-12 | Methods for predicting future technological knowledge flows |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113989075A CN113989075A (en) | 2022-01-28 |
| CN113989075B true CN113989075B (en) | 2025-01-28 |
Family
ID=79738259
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111186872.8A Active CN113989075B (en) | 2021-10-12 | 2021-10-12 | Methods for predicting future technological knowledge flows |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113989075B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114579749B (en) * | 2022-03-11 | 2024-08-23 | 北京中知智慧科技有限公司 | CPC classification method and device for patent data |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102711914A (en) * | 2009-12-28 | 2012-10-03 | 皇家飞利浦电子股份有限公司 | High intensity focused ultrasound transducer optimization |
| CN108351879A (en) * | 2015-07-16 | 2018-07-31 | 构造数据有限责任公司 | The system and method for the division search index of efficiency for improving identification media segment |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH11241992A (en) * | 1997-09-10 | 1999-09-07 | Kiyoaki Kobayashi | Spectrophotometric measurement of mobile object and method and apparatus for estimation of volume ratio |
| CN101346743A (en) * | 2005-12-29 | 2009-01-14 | 卡尔斯特里姆保健公司 | Cross-time and cross-modality medical diagnosis |
| EP2526500A2 (en) * | 2010-01-18 | 2012-11-28 | Elminda Ltd. | Method and system for weighted analysis of neurophysiological data |
| JP6580567B2 (en) * | 2013-12-04 | 2019-09-25 | オバロン・セラピューティクス、インコーポレイテッドＯｂａｌｏｎ Ｔｈｅｒａｐｅｕｔｉｃｓ， Ｉｎｃ． | Systems and methods for deploying and / or characterizing intragastric devices |
| US11138827B2 (en) * | 2016-09-15 | 2021-10-05 | Simpsx Technologies Llc | Implementations of a computerized business transaction exchange for various users |
| CN110428010B (en) * | 2019-08-05 | 2022-09-06 | 中国科学技术大学 | Knowledge tracking method |
| CN110580261B (en) * | 2019-09-09 | 2022-07-15 | 中国科学技术大学 | An in-depth technology tracking method for high-tech companies |
| CN112182308B (en) * | 2020-09-29 | 2023-03-24 | 华中师范大学 | Multi-feature fusion depth knowledge tracking method and system based on multi-thermal coding |
-
2021
- 2021-10-12 CN CN202111186872.8A patent/CN113989075B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102711914A (en) * | 2009-12-28 | 2012-10-03 | 皇家飞利浦电子股份有限公司 | High intensity focused ultrasound transducer optimization |
| CN108351879A (en) * | 2015-07-16 | 2018-07-31 | 构造数据有限责任公司 | The system and method for the division search index of efficiency for improving identification media segment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113989075A (en) | 2022-01-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yu et al. | Evolving artificial neural networks using an improved PSO and DPSO | |
| CN104270281B (en) | A Combination Prediction Method for Web Service QoS | |
| CN111125358B (en) | Text classification method based on hypergraph | |
| Labonne et al. | Linking dendritic network structures to population demogenetics: the downside of connectivity | |
| Xiang et al. | A clustering-based surrogate-assisted multiobjective evolutionary algorithm for shelter location problem under uncertainty of road networks | |
| CN112508085A (en) | Social network link prediction method based on perceptual neural network | |
| Shambour et al. | Modified global flower pollination algorithm and its application for optimization problems | |
| Sheikholeslami et al. | A hybrid cuckoo–harmony search algorithm for optimal design of water distribution systems | |
| CN106570597A (en) | Content popularity prediction method based on depth learning under SDN architecture | |
| Wang et al. | Study on MPGA‐BP of gravity dam deformation prediction | |
| CN105184415B (en) | A kind of Distribution Networks Reconfiguration design method | |
| CN112836876B (en) | Power distribution network line load prediction method based on deep learning | |
| CN103400190A (en) | Integrated framework method for optimizing extremity learning machine by using genetic algorithm | |
| Gu et al. | [Retracted] Application of Fuzzy Decision Tree Algorithm Based on Mobile Computing in Sports Fitness Member Management | |
| CN114584406B (en) | Industrial big data privacy protection system and method for federated learning | |
| CN106528428A (en) | Method for constructing software variability prediction model | |
| CN112269729A (en) | Load intelligence analysis method for large-scale server cluster of online shopping platform | |
| CN113989075B (en) | Methods for predicting future technological knowledge flows | |
| Chen et al. | A novel neural network based on quantum computing | |
| CN114676332A (en) | Network API recommendation method facing developers | |
| CN118540166A (en) | Honey pot network security situation prediction method and system | |
| Wang | Analysis of bank credit risk evaluation model based on BP neural network | |
| CN106453294A (en) | Security situation prediction method based on niche technology with fuzzy elimination mechanism | |
| Tospornsampan et al. | Optimization of a multiple reservoir system using a simulated annealing--A case study in the Mae Klong system, Thailand | |
| Wang et al. | Applying optimized BPN to a chaotic time series problem |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |