CN114037284B - Method for predicting App popularity evolution result based on multi-layer attribute network - Google Patents
Method for predicting App popularity evolution result based on multi-layer attribute network Download PDFInfo
- Publication number
- CN114037284B CN114037284B CN202111329169.8A CN202111329169A CN114037284B CN 114037284 B CN114037284 B CN 114037284B CN 202111329169 A CN202111329169 A CN 202111329169A CN 114037284 B CN114037284 B CN 114037284B
- Authority
- CN
- China
- Prior art keywords
- version
- app
- layer
- node
- data
- 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 20
- 230000008451 emotion Effects 0.000 claims abstract description 7
- PJVWKTKQMONHTI-UHFFFAOYSA-N warfarin Chemical compound OC=1C2=CC=CC=C2OC(=O)C=1C(CC(=O)C)C1=CC=CC=C1 PJVWKTKQMONHTI-UHFFFAOYSA-N 0.000 claims abstract description 7
- 230000009193 crawling Effects 0.000 claims abstract description 5
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 36
- 238000004140 cleaning Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 5
- 230000002996 emotional effect Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 2
- 238000011160 research Methods 0.000 description 8
- 238000011161 development Methods 0.000 description 7
- 238000012423 maintenance Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012937 correction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/70—Software maintenance or management
- G06F8/71—Version control; Configuration management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/70—Software maintenance or management
- G06F8/75—Structural analysis for program understanding
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Computer Security & Cryptography (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Information Transfer Between Computers (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a method for predicting an App popularity evolution result based on a multi-layer attribute network, which comprises the following steps: data crawling; preprocessing data; constructing an App multi-layer attribute network; step four, constructing a user popularity prediction model: calculating the similarity of the Code layer network by using DeltaCon algorithm according to the App multi-layer attribute network to obtain the comparison result of the latest version of each version; and taking the emotion tendency data contained in the Tag layer of each version as the popularity of the user, and finally calculating the prediction result of the popularity of the user of the latest version through a prediction method.
Description
Technical Field
The invention relates to the field of software engineering technology and application, in particular to a method for predicting android App popularity evolution results based on a multi-layer attribute network.
Background
The evolution of software plays an increasingly important role in software development. Programmers rarely build software from scratch, but typically spend more time modifying existing software, providing new functionality to users and repairing defects in existing software. Evolving software systems are typically a time consuming and error prone process. In fact, it is reported that 90% of the cost of a typical software system occurs during the maintenance phase, and a major focus of software engineering involves problems associated with upgrading, migrating and evolving existing software systems. The term software evolution dates back to 1976, when Belady and Lehman created this term for the first time. Software evolution refers to the dynamic behavior of software systems because they are maintained and enhanced over their lifecycle. Software evolution is particularly important as systems in organizations become longer lived.
With the rapid development of the mobile internet, software engineering has changed greatly. Open source is a major form of software technology innovation and industry development. The development of software is increasingly being focused on by modern software engineering and is becoming more and more focused. Compared with the traditional software, the mobile phone application program has many new characteristics in development mode and is quite different. But most of the evolution research of the existing software version is focused on applications made of desktop programming language, so far there is little research on evolution of App version based on mobile terminal. With the popularity of mobile apps, maintenance of developers has become very important due to the intense competition in the application market. Ensuring that the application program is continuously developed and works normally. In order to improve the quality of an application, developers must continually update the version of the application. To reduce the production and development costs of the application program. The application manager must know the factors that evolve the application. Knowing the evolution of the application functionality may help the developer reduce maintenance costs and increase the utilization possibilities of the application.
In recent years, metadata such as open applications and software history correction information have been stored in large amounts. And a powerful support is provided for maintenance management of software project development. An open source software platform such as Github provides many references for the next experiment. In 2004, the generation mode of network information is changed over the sky, and batch uploading data replaces the data generated by the traditional network platform, namely crowd-sourced data, namely data contributed by a user group through various participation modes. Users of application stores such as Google Play can comment on their own use, set preferences, etc., and can score and evaluate applications as well, which are part of a typical intelligence community created by a large number of application users, and data is collected and analyzed to help learn about the evolution of application attributes.
The prior art has at least the following disadvantages:
(1) The evolution research of the currently existing software versions is mostly focused on the traditional application programs, but little research is done on the version development of the mobile terminal App. The existing software evolution research can not well reflect the actual situation of application program development.
(2) The existing evolution research has single visual angle. Most research focused on the how argument of software evolution and source code changes inside the software, with less attention paid to what and why views of software evolution. That is, there is a lack of research into interactions with software users.
Aiming at the problems, the invention provides a method for predicting the user popularity of a newly released version of an Android App. And collecting the source code change and the theme of crowd-sourced data in the evolution process, predicting the popularity of the new version based on the similarity of the graphs, and reducing the cost of software maintenance.
Disclosure of Invention
The invention aims to provide a method for predicting an App popularity evolution result. Under the framework, a user provides a mobile phone application program App needing to be predicted, the system firstly climbs crowd data such as App source Code change history, user comments in an application market and the like, constructs a multi-layer attribute network comprising a Code layer, an App layer and a Tag layer for each version, and finally outputs a user popularity prediction result of the latest version according to the similarity of the attribute network of each version, thereby saving maintenance time and cost. The technical proposal is as follows:
A method for predicting an App popularity evolution result based on a multi-layer attribute network comprises the following steps:
stage one data crawling: the evolution history data is crawled through the open source software version control system application programming interface GithubApi and user comments are crawled through the user Token of the android mobile software application market GooglePlay.
Stage two data preprocessing: and comprehensively considering user comments in GooglePlay of the source code historical data in Github, extracting enough corresponding versions of the user comments and other data, and cleaning the data of the crawled natural language.
Step three, constructing an App multi-layer attribute network: generating a control flow graph of each function in the application program according to the source Code data, extracting calling information and called information from the issuing calling node, and finally generating a function calling graph of a Code layer; according to the source code data, obtaining the document similarity of two adjacent versions, extracting the theme of each document by the LDA, obtaining the theme set of the version code, and finally forming an App layer network with the metadata of the version; and extracting labels of the user comments and emotion tendencies corresponding to the comments by using LDA according to data such as the user comments in GooglePlay, and finally mapping the labels to a WordNet network to generate a Tag layer network.
Step four, constructing a user popularity prediction model: calculating the similarity of the Code layer network by using DeltaCon algorithm according to the App multi-layer attribute network to obtain the comparison result of the latest version of each version; and taking the emotion tendency data contained in the Tag layer of each version as the popularity of the user, and finally calculating the prediction result of the popularity of the user of the latest version through a prediction method.
Preferably, the specific steps of constructing the App multi-layer attribute network in the stage three are as follows:
(1) Constructing a Code layer: generating a function call graph on the basis of generating a control flow graph CFG, wherein each node in the CFG of each function corresponds to each statement in the program, and each node object has attributes, including whether the node is a branch judgment node, whether the node is a called node or not, and whether the node is an exit node or not; each node will record which function it belongs to. Traversing the nodes in the CFG can obtain the relation between the call and the called nodes, namely, find the function call nodes, extract the information of the call function and the called function, store the called function nodes and the call function nodes in the adjacent linked list, and generate the function call graph.
(2) Constructing an App layer: for the APP layer, metadata information such as version numbers, release time and the like can be grabbed from a network during data acquisition, and software features of each version are extracted from codes updated in an increment manner in github; comparing the version with the source code of the previous version, using the Levenshtein distance to determine if the source code documents of the two versions are similar, if the two documents are more than 0.98 similar, they are considered duplicate documents, and deleting them. Subject extraction is performed on the remaining non-duplicate documents to obtain the software features of the version and to wash away many of the programming language grammar related characters and keywords contained in the documents. The LDA is then used to extract the keywords of the source code from the entity identifier and code notes of each document, the collection of which constitutes the software features of each version.
(3) Constructing a Tag layer: the LDA model is implemented using python's gensim package, which extracts the subject terms for each comment, and the subject terms extracted from all comments are over-weighted to form a tag set. For each version of the app, the resulting set of labels, each label is restored to a meta-word by the set of synonyms in wordnet, which can all find the corresponding mapping node in wordnet. And obtaining a label network corresponding to each version of the app through the node matching of the labels and wordnet.
Preferably, stage four builds a user popularity prediction model: the method comprises the following specific steps:
For two versions of the app, the similarity of the function call graphs of the two versions is calculated, and the similarity of the graphs is calculated using DeltaCon algorithm. The algorithm mainly calculates the similarity between two graphs by comparing the connectivity of the same nodes in the two graphs, utilizes the code layer in the app multi-layer attribute network, firstly calculates the influence degree of paired nodes in the code layer by using a confidence propagation algorithm, generates a code layer influence degree matrix of two adjacent versions, then calculates the root Euclidean distance of the two matrixes, determines the difference of the influence scores of the same nodes in the two graphs, and finally integrates the difference into the similarity score of the code layer of the two versions;
For all versions of the app, calculating similarity scores of the app and the latest version, wherein in a multi-layer attribute network of the app, tag layers contain keywords extracted from user comments of the application market, the tags are marked with respective emotional tendencies, and the duty ratio of forward tags is counted to serve as the popularity of the app version. And calculating the product of the similarity ratio of each version and the latest version and the popularity, and normalizing to obtain the predicted value of the user popularity of the App of the latest version.
Drawings
FIG. 1 is a diagram of an overall framework for a new version of user popularity prediction in accordance with the present invention.
Fig. 2 is a list of test uses App according to the present invention.
Fig. 3 is a first layer in the App attribute network of the present invention.
Fig. 4 shows a third layer in the App attribute network according to the present invention.
FIG. 5 is a schematic diagram of predicting user popularity according to the present invention.
Detailed Description
In order to overcome the defects of the prior art, the invention aims to provide a prediction method of an android App popularity evolution result. Under the framework, a user provides a mobile phone application program App needing to be predicted, the system firstly climbs crowd data such as App source Code change history, user comments in an application market and the like, constructs a multi-layer attribute network comprising a Code layer, an App layer and a Tag layer for each version, and finally outputs a user popularity prediction result of the latest version according to the similarity of the attribute network of each version, thereby saving maintenance time and cost.
The specific execution flow of the method is as follows:
stage one data crawling: the evolution history data is crawled through the open source software version control system application programming interface GithubApi and user comments are crawled through the user Token of the android mobile software application market GooglePlay.
Stage two data preprocessing: for the source code historical data in Github, comprehensively considering the user comments in GooglePlay, and extracting corresponding versions with enough information such as the user comments. And cleaning the data of the crawled natural language.
Step three, constructing an App multi-layer attribute network: generating a control flow graph of each function in the application program according to the source Code data, extracting calling information and called information from the function calling node, and finally generating a function calling graph of a Code layer; according to the source code data, obtaining the document similarity of two adjacent versions, extracting the theme of each document by the LDA, obtaining the theme set of the version code, and finally forming an App layer network with the metadata of the version; according to the crowd data in GooglePlay, the LDA extracts the label of the user comment and the emotion tendency corresponding to the comment, and finally maps the label to the WordNet network to generate a Tag layer network.
Step four, constructing a user popularity prediction model: calculating the similarity of the Code layer network by using DeltaCon algorithm according to the App multi-layer attribute network to obtain the comparison result of the latest version of each version; and taking emotion tendency data contained in the Tag layer of each version as user popularity, and finally calculating a prediction result of the user popularity of the latest version through a prediction algorithm model.
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the embodiments of the present invention will be described in further detail with reference to the accompanying drawings.
Step one data collection
The data collection is the first main stage, 8 mobile terminal application items made in Android languages are selected from Github, and required source codes and historical submission information are acquired from each Github storage device. For these open source APPs that were up-shelved in GooglePlay, annotation data and scoring information for all users is crawled. Experiments were performed on Android App in 8, the selected App being shown in fig. 2.
Data crawling:
(1) With respect to each application of the experimental dataset, a github user's personal access token was obtained following similar metadata information and source code version history acquisition steps. Then, through GRAPHQL API of github, all tags of historical versions of an app are obtained, graphQL is an interface query language pushed by Facebook, required data can be defined correctly, redundancy is avoided, and through GraphQL, data obtained by a plurality of REST requests can be obtained only by one request. And according to the tag names of the versions, the modification information and the source code downloading address of each version can be obtained.
(2) To obtain GooglePlay comments of all users, the address of Google Play of the application is first obtained. Information such as comments of the user is sent to the front end in the format of json data, so that the comments of the user can be obtained through the homepage address of the application. By GoolePlay settings, at most 200 comments can be displayed per page. The json data of comments of all users are required to be grabbed through the retrieval of each page.
Data preprocessing:
(1) For the data of user comments in Goolge Play, data cleaning is required, because the comment data at this time includes other languages such as English, chinese, and the like. Because text analysis is performed in english, comments other than english need to be filtered. This step uses LANG DETECT for language checking.
(2) After filtering non-english comments, special characters such as chinese characters, punctuation marks, expressions and the like are mixed in the english comments. The minimum unit granularity of data analysis is desirably a word. For english corpus, the smallest unit of english sentence is a word. Here, a tokenization operation is performed using a word segmenter of nltk toolkit.
(3) Stop term stopword refers to a term that does not have any co-attractive effect on the study goal. Such as Hello, am, is, etc., also have punctuation marks. These are not intended to be imported at the time of text analysis, and therefore need to be deleted.
Step two, generating an App multi-layer attribute network
(1) Constructing a Code layer: a function call graph is generated on the basis of generating a Control Flow Graph (CFG), and each node in the CFG of each function corresponds to each statement in the program (assuming that each statement code occupies one line). Each node object has some attributes, such as whether it is a branch decision node, a called node, or an exit node. While each node records which method it belongs to. Therefore, the nodes in the CFG are traversed to obtain the relation between the call and the called nodes, namely, the function call node is found, and the information of the call function and the called function is extracted. And storing the called function node and the calling function node in an adjacent linked list to generate a function call graph. See fig. 3.
(2) Constructing an App layer: for the APP layer, metadata information such as version numbers, release time and the like can be grabbed from the network during data acquisition. The software features of each version are extracted from the code incrementally updated in github, the version is compared with the source code of the previous version, the Levenshtein distance is used to determine if the source code documents of the two versions are similar, if the similarity of the two documents exceeds 0.98, they are considered duplicate documents, and they are deleted. Subject extraction is performed on the remaining non-duplicate documents to obtain the software features of the version and to wash away many of the programming language grammar related characters and keywords contained in the documents. The LDA is then used to extract the keywords of the source code from the entity identifier and code notes of each document, the collection of which constitutes the software features of each version.
(3) Constructing a Tag layer: the LDA model is implemented using python's gensim package, three subject words are extracted for each comment, and the subject words extracted from all comments are repeated to form a tag set. For each version of the app, the resulting set of labels, each label is restored to a meta-word by the set of synonyms in wordnet, which can all find the corresponding mapping node in wordnet. By simple label matching with wordnet nodes, see fig. 4, a label network corresponding to each version of the app can be obtained.
Step three, predicting the popularity of the new version
For two versions of the app, the similarity of the function call graphs of the two versions is calculated, see fig. 5. The similarity of the graphs is calculated here using the DeltaCon algorithm. The algorithm calculates the similarity between two graphs, mainly by comparing the connectivity of the same nodes in the two graphs. The method comprises the steps of utilizing a code layer in an app multi-layer attribute network, firstly calculating influence degrees of paired nodes in the code layer by using a confidence propagation algorithm, generating code layer influence degree matrixes of two adjacent versions, then calculating root Euclidean distance of the two matrixes, determining difference of influence scores of the same nodes in the two graphs, and finally integrating the difference into similarity scores of the code layers of the two versions.
For all versions of the app, its similarity score to the latest version is calculated. In the multi-layer attribute network of the app, the tag layer contains keywords extracted from user comments of the application market, the tags are also marked with respective emotional tendencies, and the duty ratio of the forward tags needs to be counted as the popularity of the app version. And calculating the product of the similarity ratio of each version and the latest version and the popularity, and normalizing to obtain the predicted value of the user popularity of the App of the latest version.
Claims (1)
1. A method for predicting an App popularity evolution result based on a multi-layer attribute network comprises the following steps:
Stage one data crawling: the evolution history data is crawled through an application programming interface GithubApi of the open source software version control system, and user comments are crawled through a user Token of the android mobile software application market GooglePlay;
Stage two data preprocessing: comprehensively considering user comments in GooglePlay of source code historical data in Github, extracting enough corresponding versions of the user comment data, and cleaning the data of the crawled natural language;
Step three, constructing an App multi-layer attribute network: generating a control flow graph of each function in the application program according to the source Code data, extracting calling information and called information from the issuing calling node, and finally generating a function calling graph of a Code layer; according to the source code data, obtaining the document similarity of two adjacent versions, extracting the theme of each document by the LDA, obtaining the theme set of the version code, and finally forming an App layer network with the metadata of the version; according to user comment data in GooglePlay, extracting a label of a user comment and emotion tendencies corresponding to the comment by using LDA, and finally mapping the label into a WordNet network to generate a Tag layer network, wherein the method comprises the following steps:
(1) Constructing a Code layer: generating a function call graph on the basis of generating a control flow graph CFG, wherein each node in the CFG of each function corresponds to each statement in the program, and each node object has attributes, including whether the node is a branch judgment node, whether the node is a called node or not, and whether the node is an exit node or not; each node records which function it belongs to; traversing the nodes in the CFG to obtain the relation between the call and the called, namely finding out the function call node, extracting the information of the call function and the called function, storing the called function node and the call function node in an adjacent linked list, and generating a function call graph;
(2) Constructing an App layer: for the APP layer, the version number and the release time metadata information are captured from the network during data acquisition, and the software features of each version are extracted from codes updated in an increment manner in github; comparing the version with the source code of the previous version, judging whether the source code documents of the two versions are similar or not by using the Levenshtein distance, if the similarity of the two documents exceeds 0.98, considering the two documents as duplicate documents, and deleting the duplicate documents; extracting the theme of the rest non-repeated document to obtain the software characteristics of the version, and cleaning out a plurality of characters and keywords which are contained in the document and are related to programming language grammar; extracting keywords of the source code from the entity identifier and the code annotation of each document by using the LDA, wherein the collection of the code keywords forms the software characteristic of each version;
(3) Constructing a Tag layer: using python's gensim package to implement LDA model, extracting subject words from every comment, and forming tag set after the subject words extracted from all comments are over-weighted; for the tag set obtained for each version of the app, reducing each tag to metawords by the synonym set in wordnet, wherein the metawords can find corresponding mapping nodes in wordnet; obtaining a label network corresponding to each version of the app through node matching of labels and wordnet;
Step four, constructing a user popularity prediction model: calculating the similarity of the Code layer network by using DeltaCon algorithm according to the App multi-layer attribute network to obtain the comparison result of the latest version of each version; taking emotion tendency data contained in the Tag layer of each version as user popularity, and finally calculating a prediction result of the user popularity of the latest version by a prediction method, wherein the method comprises the following steps:
For two versions of the app, calculating the similarity of the function call graphs of the two versions, and calculating the similarity of the graphs by using DeltaCon algorithm; the algorithm calculates the similarity between two graphs by comparing the connectivity of the same nodes in the two graphs, utilizes the code layer in the app multi-layer attribute network, firstly calculates the influence degree of paired nodes in the code layer by using a confidence propagation algorithm, generates a code layer influence degree matrix of two adjacent versions, then calculates the root Euclidean distance of the two matrixes, determines the difference of the influence scores of the same nodes in the two graphs, and finally integrates the difference into the similarity score of the code layer of the two versions;
for all versions of the app, calculating similarity scores of the app and the latest version, wherein in a multi-layer attribute network of the app, tag layers contain keywords extracted from user comments of the application market, the tags are marked with respective emotional tendencies, and the duty ratio of forward tags is counted to serve as the popularity of the app version; and calculating the product of the similarity ratio of each version and the latest version and the popularity, and normalizing to obtain the predicted value of the user popularity of the App of the latest version.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111329169.8A CN114037284B (en) | 2021-11-10 | 2021-11-10 | Method for predicting App popularity evolution result based on multi-layer attribute network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111329169.8A CN114037284B (en) | 2021-11-10 | 2021-11-10 | Method for predicting App popularity evolution result based on multi-layer attribute network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114037284A CN114037284A (en) | 2022-02-11 |
| CN114037284B true CN114037284B (en) | 2024-05-28 |
Family
ID=80143898
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111329169.8A Active CN114037284B (en) | 2021-11-10 | 2021-11-10 | Method for predicting App popularity evolution result based on multi-layer attribute network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114037284B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104916004A (en) * | 2014-03-13 | 2015-09-16 | 通用汽车环球科技运作有限责任公司 | Method and apparatus of tracking and predicting usage tread of in-vehicle apps |
| CN108614778A (en) * | 2018-05-10 | 2018-10-02 | 天津大学 | Prediction technique is changed based on the Android App program evolutions that Gaussian process returns |
| CN109903087A (en) * | 2019-02-13 | 2019-06-18 | 广州视源电子科技股份有限公司 | The method, apparatus and storage medium of Behavior-based control feature prediction user property value |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10389677B2 (en) * | 2016-12-23 | 2019-08-20 | International Business Machines Corporation | Analyzing messages in social networks |
-
2021
- 2021-11-10 CN CN202111329169.8A patent/CN114037284B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104916004A (en) * | 2014-03-13 | 2015-09-16 | 通用汽车环球科技运作有限责任公司 | Method and apparatus of tracking and predicting usage tread of in-vehicle apps |
| CN108614778A (en) * | 2018-05-10 | 2018-10-02 | 天津大学 | Prediction technique is changed based on the Android App program evolutions that Gaussian process returns |
| CN109903087A (en) * | 2019-02-13 | 2019-06-18 | 广州视源电子科技股份有限公司 | The method, apparatus and storage medium of Behavior-based control feature prediction user property value |
Non-Patent Citations (2)
| Title |
|---|
| 何珊珊.基于机器学习的视频软件受欢迎度研究.全文. * |
| 郝有强.移动App的潜在网络结构分析与演化分析.全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114037284A (en) | 2022-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10592505B2 (en) | Natural language interfaces for databases using autonomous agents and thesauri | |
| US11442932B2 (en) | Mapping natural language to queries using a query grammar | |
| US11790006B2 (en) | Natural language question answering systems | |
| US9864741B2 (en) | Automated collective term and phrase index | |
| US10558754B2 (en) | Method and system for automating training of named entity recognition in natural language processing | |
| US10956433B2 (en) | Performing an operation relative to tabular data based upon voice input | |
| WO2020108063A1 (en) | Feature word determining method, apparatus, and server | |
| US20120089394A1 (en) | Visual Display of Semantic Information | |
| US20160103837A1 (en) | System for, and method of, ranking search results obtained by searching a body of data records | |
| US10783877B2 (en) | Word clustering and categorization | |
| US20160103885A1 (en) | System for, and method of, building a taxonomy | |
| US11907671B2 (en) | Role labeling method, electronic device and storage medium | |
| US11281864B2 (en) | Dependency graph based natural language processing | |
| US20130124194A1 (en) | Systems and methods for manipulating data using natural language commands | |
| US20160103920A1 (en) | System for, and method of, searching data records | |
| US10977155B1 (en) | System for providing autonomous discovery of field or navigation constraints | |
| US20220129448A1 (en) | Intelligent dialogue method and apparatus, and storage medium | |
| CN112417090A (en) | Using uncommitted user input data to improve task performance | |
| US20190005028A1 (en) | Systems, methods, and computer-readable medium for validation of idiomatic expressions | |
| JP2022024102A (en) | Method for training search model, method for searching target object and device therefor | |
| US11921764B2 (en) | Utilizing artificial intelligence models to manage and extract knowledge for an application or a system | |
| CN114579104A (en) | Data analysis scene generation method, device, equipment and storage medium | |
| US20220198358A1 (en) | Method for generating user interest profile, electronic device and storage medium | |
| CA3099201A1 (en) | Emoji recommendation system and method | |
| US10534788B2 (en) | Automatically determining a recommended set of actions from operational data |
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 |