Software information processing method based on big data

Abstract

The invention discloses a software information processing method based on big data, and relates to the technical field of software information processing, comprising the following information processing method: collecting behavior logs on social media, video websites and news applications through a collection module to obtain the number of likes, the number of comments, the number of shares and the online login duration of a user; according to the method, multiple dimensions of user interaction data can be comprehensively considered, user behavior characteristics can be mined more accurately, a solid foundation is provided for subsequent preference prediction and recommendation score calculation, accurate prediction of user content preference can be realized, the accuracy of recommended content and user satisfaction are improved, and the user experience is improved. And the timeliness of the content and the matching degree between the user and the content type are considered, so that the content recommendation score better meeting the user requirement can be calculated, and the recommendation effect is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of software information processing, and particularly relates to a software information processing method based on big data. BACKGROUND

[0002] With the rapid development of Internet technology and the wide popularity of social media, social networks have become an indispensable information exchange platform in people's daily life. On the platform, user behavior data such as likes, comments and shares are continuously generated and accumulated. These data contain rich valuable information such as user preferences, interests, social relationships and consumption habits. Software information processing technology based on big data has been widely applied in many fields, especially in the field of social network analysis. It can help enterprises better understand user behavior and predict user preferences, thereby providing more personalized and accurate services.

[0003] However, the existing software information processing method based on big data may not fully consider the comprehensive interaction intensity of users with different social types of content, the influence of comprehensive interaction types such as likes, comments and shares and user activity on user behavior characteristics, resulting in inaccurate and comprehensive user behavior characteristics. When predicting user content preferences, only the direct association between users and content is considered, while the content popularity, the association between users and content types and the freshness of content are ignored, resulting in inaccurate prediction results and difficulty in meeting user personalized needs. SUMMARY

[0004] The purpose of the present application is to provide a software information processing method based on big data, which solves the problems raised in the background art.

[0005] To achieve the above purpose, the present application provides the following technical scheme: a software information processing method based on big data, comprising the following information processing method: Step I: collecting behavior logs on social media, video websites and news applications through a collection module to obtain the number of likes, the number of comments, the number of shares and the login online duration of the user; Step II: performing data cleaning, data integration and normalization processing on the number of likes, the number of comments, the number of shares and the login online duration of the user through a data processing module, and the data processing module outputs the number of likes of user i on content type j, the number of comments of user i on content type j, the number of shares of user i on content type j and the activity of user i; Step III, input the number of likes of user i to content type j, the number of comments of user i to content type j, the number of shares of user i to content type j and the activity of user i into the information processing module, and the information processing module outputs the behavior characteristic value of user i to content type j, the extended preference prediction value of user i to content type j and the recommendation score of user i to content type j; Step IV, analyze the behavior characteristic value of user i to content type j, the extended preference prediction value of user i to content type j and the recommendation score of user i to content type j by the analysis module to deeply understand the user demand and preference and recommend the content most suitable for the preference of the user.

[0006] Optionally, the information processing module comprises a user behavior submodule, a content preference submodule, a recommendation analysis submodule and a correlation degree submodule.

[0007] Optionally, the calculation formula of the user behavior submodule is as follows: UPSA j，i =(LKA j，i ×SA+CON j，i ×SB+SAE j，i ×SC) 2 +ACCTY i -QLT j -0.5 ; Wherein: UPSA j，i represents the behavior characteristic value of user i to content type j; j represents content type, which is different types of content in the field of social network analysis, and the content type is text, picture, video and audio, LKA j，i represents the number of likes of user i to content type j, CON j，i represents the number of comments of user i to content type j, SAE j，i表示 represents the number of shares of user i to content type j, ACCTY i represents the activity of user i, QLT j represents the quality of content type j, SA represents the weight coefficient of likes, SB represents the weight coefficient of comments, and SC represents the weight coefficient of shares. The processing process of the user behavior submodule is as follows: input the number of likes LKA j，i of user i to content type j, the number of comments CON j，i of user i to content type j, the number of shares SAE j，i of user i to content type j and the activity ACCTY i of user i into the user behavior submodule, and output the behavior characteristic value UPSAj，i .

[0008] Optionally, the calculation formula of the content preference sub-module is as follows: CPAQ j，i = (UPSA j，i × PS j ) + (AFNY j，i × FRHE 0.5 ) - (DTET j，i / PLAT i ); Wherein: CPAQ j，i represents the extended preference prediction value of the user i to the content type j; PS j represents the popularity of the content type j, AFNY j，i represents the association degree of the user i to the content type j, FRHE represents the freshness of the content type j, DTET j，i represents the degree of disinterest of the user i to the content type j, PLAT i represents the popularity of the user i, UPSA j，i × PS j represents the weighted effect of the user behavior characteristics on the popular content; The processing process of the content preference sub-module is as follows: the user i behavior characteristic value UPSA j，i of the user i to the content type j is input to the content preference sub-module, and based on the popularity PS j of the content type j and the association degree AFNY j，i of the user i to the content type j, the extended preference prediction value CPAQ j，i of the user i to the content type j is output.

[0009] Optionally, the calculation formula of the recommendation analysis sub-module is as follows: CRSQ j，i = (CPAQ j，i × RCN j ) + (UPSA j，i × TA) + (MTC 0.5 × PPS); Wherein: CRSQ j，i represents the recommendation score of the user i to the content type j; RCN j represents the timeliness of the content type j, TA represents the weight factor of UPSA j，i , MTC represents the matching degree of the user i to the content type j, and PPS represents the weight factor of MTC; The processing procedure of the recommendation analysis submodule is as follows: the extended preference prediction value CP AQ of the user i to the content type j j，i and the behavior characteristic value UP SA of the user i to the content type j j，i are input into the recommendation analysis submodule, and the recency RC N of the content type j is used to output the recommendation score CR SQ of the user i to the content type j j .

[0010] Optionally, the processing procedure of the correlation submodule is as follows: Step one: AF NY j，i，new = AF NY j，i，old + a x (CR SQ j，i - T SD) ; Step two: establish a loop termination condition; Condition one: |AF NY j，i，new - AF NY j，i，old | < 0.00493; Condition two: the iteration number is 123 times; wherein: AF NY j，i，new represents the correlation of the user i to the content type j after iteration; AF NY j，i，old represents the correlation of the user i to the content type j before iteration, a represents a learning rate, used to control the step length of updating, T SD represents a comparison threshold, used to determine whether CR SQ j，i reaches an expected level; Based on the difference between the recommendation score CR SQ j，i of the user i to the content type j and the comparison threshold T SD, and by using the learning rate a to control the change degree of the correlation of the user i to the content type j before iteration AF NY j，i，old , the correlation of the user i to the content type j after iteration AF NY j，i，new is outputted. j，i，new The correlation of the user i to the content type j after iteration AF NY j，i is used to replace the correlation of the user i to the content type j in the content preference submodule AF NY j，i , so as to continuously optimize CP AQ j，i and CR SQ j，i , and through condition one and condition two, the loop termination convergence is achieved.

[0011] Optionally, the collection module can obtain user behavior data through web crawler technology, obtain user data through log analysis software, and obtain user behavior data through API interface calling tools.

[0012] Optionally, the data cleaning in the data processing module is to remove duplicate data and correct error data, the data integration is to integrate data from different sources, and the data is standardized to ensure the consistency of the data format, the normalization processing is to convert the data into a format suitable for processing, and then the data is normalized to ensure that the data is on the same scale.

[0013] Compared with the prior art, the present application has the following advantages: I. The present application outputs the behavior characteristic value of user i to content type j UPSA j，i The behavior sub-module can comprehensively consider various behavior characteristics of the user, such as the number of views, the number of clicks, the dwell time and the number of shares, and can further comprehensively reflect the interests and preferences of the user, and multi-dimensional analysis makes the user portrait more stereoscopic and accurate, UPSA j，i The calculation provides basic data for subsequent personalized recommendation, and the in-depth analysis of the behavior characteristics of the user sub-module can identify the unique interests and needs of the user, and provide strong support for personalized recommendation, UPSA j，i Not only the basic attributes of the user are considered, but also the behavior characteristics of the user are analyzed in depth, so that the user portrait is more comprehensive and accurate, and the user sub-module can also dynamically reflect the changes of user interest, thereby improving the timeliness and accuracy of the recommendation.

[0014] II. The present application outputs the extended preference prediction value of user i to content type j CPAQ j，i The user sub-module considers multiple factors such as user behavior characteristics, content popularity, and user content type correlation, thereby accurately predicting user preference for specific content, and the user sub-module can recommend content that is more in line with the interests and needs of the user while improving the diversity of the recommendation, CPAQ j，i The calculation can provide the basis for the optimization strategy of the recommendation system, and the recommendation system can adjust the recommendation strategy through the analysis of the user preference prediction result to improve the accuracy and efficiency of the recommendation, CPAQ j，i The calculation considers more factors including user behavior characteristics and content popularity, so that the recommendation result is more comprehensive and accurate, and in addition, the user sub-module can also deeply mine the potential interests behind the user behavior, thereby improving the diversity and innovation of the recommendation.

[0015] III. The present application outputs the recommendation score of user i to content type j CRSQ j，i CPAQ j，i Based on the user preference prediction result, the user sub-module can further calculate the recommendation score of the user to specific content, which can intuitively reflect the interest degree of the user to the content and provide accurate scoring basis for the recommendation system, and the user sub-module provides more accurate scoring basis for the recommendation system, so that the recommendation result is more in line with the expectations and needs of the user, CPAQj，i The calculation can provide basis for optimization and improvement of the recommendation system, and the recommendation system can find potential problems and improvement direction through analyzing the recommendation score distribution, thereby improving the accuracy of the recommendation, the sub-module considers more factors including user behavior characteristics, content preference prediction, etc., so that the scoring result is more comprehensive and accurate, in addition, the sub-module can also comprehensively consider the multi-dimensional relationship between the user and the content, thereby improving the objectivity and fairness of the scoring.

[0016] Four, the present application is through the correlation degree sub-module to iterate the correlation degree AFNY of user i and content type j j，i The iteration form dynamically adjusts the correlation degree of the user and the content type according to the user's behavior and the change of the content, and the dynamic nature makes the recommendation system more accurately reflect the current interest and preference of the user, thereby realizing more personalized recommendation, and through continuous iteration and optimization of AFNY j，i The recommendation system can gradually reduce the deviation between the recommendation and the actual interest of the user, thereby improving the accuracy of the recommendation, the iteration form reflects the complexity and dynamic nature of software information when processing and analyzing large-scale data, and through continuous iteration and optimization of the model parameters, the software can more accurately capture and reflect the information and mode in the data, thereby providing better service for the user. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The step flowchart of the software information processing method based on big data; Figure 2 The overall structure diagram of the software information processing method based on big data; Figure 3 The structure diagram of the information processing module of the software information processing method based on big data; Figure 4 The step flowchart of the information processing module of the software information processing method based on big data. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0019] Compared with the existing software information processing method, the existing information processing method can only collect, store and preliminarily analyze the user behavior data in the field of social network analysis, lacks in-depth mining and comprehensive understanding of the user behavior characteristics, and when predicting the user content preference, the existing technology also fails to fully consider the multi-dimensional factors such as content popularity, user and content type correlation and content freshness, resulting in inaccurate prediction results. The module of the software information processing method based on big data can more accurately mine the user behavior characteristics by comprehensively considering the multiple dimensions of the user interaction data, including the interaction times, interaction type weight, user activity and content quality, and can provide a solid foundation for subsequent content preference prediction and recommendation score calculation. By fusing the key information such as content popularity, user and content type correlation and content freshness, the accurate prediction of user content preference can be realized. When calculating the content recommendation score, the content timeliness, user and content type matching degree and other conventional factors are considered.

[0020] Embodiment one: please refer to Figures 1 to 4 The embodiment provides a software information processing method based on big data, which comprises the following information processing method: Step I, collecting the behavior logs on social media, video websites and news applications through the acquisition module to obtain the like number, comment number, share number and login online duration of the user; Step II, the data processing module performs data cleaning, data integration and normalization processing on the like number, comment number, share number and login online duration of the user, and outputs the like number of user i to content type j, the comment number of user i to content type j, the share number of user i to content type j and the activity of user i; Step III, inputting the like number of user i to content type j, the comment number of user i to content type j, the share number of user i to content type j and the activity of user i into the information processing module, and the information processing module outputs the behavior characteristic value of user i to content type j, the extended preference prediction value of user i to content type j and the recommendation score of user i to content type j; Step IV, the analysis module analyzes the behavior characteristic value of user i to content type j, the extended preference prediction value of user i to content type j and the recommendation score of user i to content type j to deeply understand the user demand and preference and recommend the content most suitable for the user preference to the user; The information processing module comprises a user behavior submodule, a content preference submodule, a recommendation analysis submodule and a correlation submodule.

[0021] In this embodiment: the mutual association and complementation of the plurality of sub-modules in the information processing module jointly constitute the core part of the software information processing method based on big data. The synergistic effect between them enables the recommendation system to more accurately understand user needs, improve the accuracy and efficiency of recommendations, and the method can provide personalized recommendation services for users. This personalized recommendation not only improves user satisfaction and loyalty, but also enhances user trust and dependence on the recommendation system. The combination of multiple sub-modules also provides a basis for the optimization and improvement of the recommendation system. By analyzing the sub-module calculation results and the recommendation effect, the recommendation system can find potential problems and improvement directions, thereby continuously improving the accuracy and efficiency of recommendations. In the era of big data, the recommendation system needs to process a large amount of user and content data. The combination of multiple sub-modules fully reflects the demand for big data processing capabilities. They can effectively utilize these data to extract valuable information for recommendation services. The combination of multiple sub-modules also reflects the technical means of software information processing, which realizes deep mining and analysis of data through algorithms and models, improving the accuracy and efficiency of recommendations. This technical means enables the recommendation system to respond more quickly to user needs and provide higher-quality recommendation services.

[0022] Please refer to Figures 1 to 4 , the user behavior sub-module processing process is as follows: UPSA j，i = (LKA j，i × SA + CON j，i × SB + SAE j，i × SC) 2 + ACCTY i - QLT j -0.5 ; Where: UPSA j，i represents the behavior characteristic value of user i to content type j; j represents content type, which is different types of content in the field of social network analysis. Content types are text, pictures, videos, and audio, LKA j，i represents the number of likes of user i to content type j, CON j，i represents the number of comments of user i to content type j, SAE j，i表示 represents the number of shares of user i to content type j; ACCTY i represents the activity of user i, which is calculated based on total interactions, online duration, access frequency, and other indicators on the platform. Users with high activity usually have higher interest and participation in platform content; QLT jrepresents the quality of content type j, the content quality is evaluated in multiple ways, the content quality is based on user interaction, that is, likes, comments and shares, the professionalism, originality, appeal of the content and other factors, high-quality content can usually attract more user attention and interaction; SA represents the weight coefficient of likes, SB represents the weight coefficient of comments, and SC represents the weight coefficient of shares. The processing process of the user behavior submodule is as follows: the number of likes LKA j，i of user i to content type j, the number of comments CON j，i of user i to content type j, the number of shares SAE j，i of user i to content type j, and the activity ACCTY i of user i are input into the user behavior submodule, and the behavior feature value UPSA j，i of user i to content type j is output.

[0023] In this embodiment: this submodule can comprehensively consider various behavior characteristics of the user, such as the number of views, the number of clicks, the dwell time, the number of shares, the comment content, etc., so as to more comprehensively reflect the interests and preferences of the user, such a multi-dimensional analysis makes the user portrait more stereoscopic and accurate, and this submodule can be dynamically updated with the change of user behavior, capturing the subtle changes in user interest, such dynamic nature enables the recommendation system to timely adjust the recommendation strategy, improving the timeliness and accuracy of the recommendation, the calculation of UPSA j，i provides basic data for subsequent personalized recommendation, by deeply analyzing the user behavior characteristics, this submodule can identify the unique interests and needs of the user, providing strong support for personalized recommendation, traditional user portrait technology based on big data software information processing often only focuses on the basic attributes of the user, such as age, gender, occupation, etc., and ignores the behavior characteristics of the user, such limitation makes the user portrait not stereoscopic and accurate, and it is difficult to meet the needs of personalized recommendation, compared with the prior art, UPSA j，i not only considers the basic attributes of the user, but also deeply analyzes the behavior characteristics of the user, making the user portrait more comprehensive and accurate, in addition, this submodule can also dynamically reflect the changes in user interest, improving the timeliness and accuracy of the recommendation, in the big data era, user behavior data is massive and complex, UBFEij can effectively process and utilize these data, extracting valuable information to provide strong support for software information processing, UPSA j，iis an important part of software information processing, which can convert user behavior data into analyzable indicators, providing the basis for subsequent content preference prediction and recommendation. This conversion process enables the recommendation system to better understand user needs, improve recommendation satisfaction and conversion rate. Subscripts i and j represent users and content types, respectively. This representation allows the sub-module to clearly express the behavior characteristics relationship between users and specific content types. The parameter subscripts i and j help accurately locate specific users and content types, improving processing accuracy and efficiency, allowing the recommendation system to respond more quickly to user needs.

[0024] Please refer to Figures 1 to 4 , the content preference sub-module processing process is as follows: CPAQ j，i = (UPSA j，i × PS j ) + (AFNY j，i × FRHE 0.5 ) - (DTET j，i / PLAT i ); Where: CPAQ j，i represents the extended preference prediction value of user i for content type j; PS j represents the popularity of content type j, PS j is calculated based on all users' interaction data with the content, reflecting the general popularity of the content. This parameter is obtained through direct data collection via API interface from the content management system or data analysis platform; AFNY j，i represents the relevance of user i and content type j, which is calculated based on the similarity between user's historical behavior and content theme, reflecting the user's interest in specific content type. It is calculated through comprehensive analysis of user historical behavior and content characteristics; FRHE represents the freshness of content type j, which is calculated based on the difference between content publishing time and current time, reflecting the timeliness and novelty of the content, by querying the timestamp information in the content management system; DTET j，i represents the degree of user i's disinterest in content type j, which is calculated based on the estimated value of user historical behavior, reflecting the user's rejection of specific content type. It is calculated through comprehensive analysis of user historical behavior; PLAT irepresents the popularity of user i, which is calculated based on the total number of interactions or followers on the social network, reflecting the user's activity and influence on the social network; UPSA j，i ×PS j represents the weighted effect of user behavior characteristics on popular content; The processing process of the content preference sub-module is as follows: the behavior characteristic value of user i for content type j UPSA j，i is input into the content preference sub-module, and based on the popularity of content type j PS j and the association degree of user i and content type j AFNY j，i outputs the extended preference prediction value of user i for content type j CPAQ j，i .

[0025] In this embodiment: this sub-module can comprehensively consider user behavior characteristics, content popularity, user association with content type and other factors, so as to more accurately predict user preference for specific content, this sub-module can recommend content that meets the interests and needs of users, while improving the diversity of recommendations, such diversity enables the recommendation system to provide users with more choices, improving user satisfaction and loyalty, the calculation of CPAQ j，i can provide the basis for optimization strategy for the recommendation system, by analyzing the results of user preference prediction, the recommendation system can adjust the recommendation strategy, improve the accuracy and efficiency of recommendation, traditional calculation and processing form often only focuses on the direct association between user and content, ignoring the potential interest behind user behavior; while the content-based recommendation algorithm only focuses on the characteristics of the content itself, ignoring the interaction behavior between user and content, both algorithms have limitations, it is difficult to meet the demand of personalized recommendation, compared with the prior art, the calculation of CPAQ j，i considers more factors, including user behavior characteristics, content popularity, etc., making the recommendation result more comprehensive and accurate, in addition, this sub-module can also deeply mine the potential interest behind user behavior, improving the diversity and innovation of recommendation, in the era of big data, content data is also massive and complex, the calculation of CPAQ j，i can effectively process and utilize these data to recommend valuable content to users, CPAQ j，i is one of the key steps in software information processing, which can combine user and content data to realize personalized content recommendation, this combination makes the recommendation system more accurately understand user needs, improve recommendation satisfaction and conversion rate, the subscripts i and j also represent users and content types, this representation makes the sub-module clearly express the preference relationship between user and specific content type.

[0026] Please refer to Figures 1 to 4, the recommended analysis submodule processing process as follows: CRSQ j，i = (CPAQ j，i × RCN j ) + (UPSA j，i × TA) + (MTC 0.5 × PPS); Wherein: CRSQ j，i represents the recommended score of user i to content type j, used to evaluate the recommended value of content to the user; RCN j represents the timeliness of content type j, which is calculated based on the difference between the content publishing time and the current time, reflecting the latest state of the content; TA represents the weight factor of UPSA j，i ; MTC represents the matching degree of user i and content type j, which is calculated based on the similarity of user interest and content theme, reflecting the adaptability and interest degree of user to content; PPS represents the weight factor of MTC; The processing process of the recommended analysis submodule is as follows: the extended preference prediction value CPAQ j，i of user i to content type j and the behavior characteristic value UPSA j，i of user i to content type j are input into the recommended analysis submodule, and the timeliness RCN j of content type j is output to the recommended score CRSQ j，i of user i to content type j.

[0027] In this embodiment: CPAQ j，i can further calculate the recommended score of user to specific content based on the result of user preference prediction, which can intuitively reflect the interest degree of user to content, providing more accurate scoring basis for recommendation system, and this submodule can provide more accurate scoring basis for recommendation system, making the recommended result more in line with the expectations and needs of user, which improves the satisfaction and conversion rate of recommendation, enhances the trust and dependence of user to recommendation system, and CPAQ j，iThe calculation of CPAQ can provide the basis for the optimization and improvement of the recommendation system. By analyzing the distribution of the recommendation score, the recommendation system can find potential problems and improvement directions, thereby improving the accuracy and efficiency of the recommendation. However, the traditional scoring system only focuses on the direct score of the content, ignoring the multi-dimensional relationship between the user and the content. This limitation makes the scoring result not comprehensive and accurate, and it is difficult to meet the demand of personalized recommendation. Compared with the prior art, the sub-module considers more factors, including user behavior characteristics, content preference prediction, etc., making the scoring result more comprehensive and accurate. In addition, the sub-module can also consider the multi-dimensional relationship between the user and the content, improving the objectivity and fairness of the scoring. In the big data era, the recommendation system needs to process a large amount of user and content data, and the calculation of CPAQ j，i can effectively utilize these data to provide personalized recommendation services for users, and the calculated CPAQ j，i is one of the important components in software information processing, which can convert the results of user preference prediction into specific recommendation scores, providing the basis for the optimization and improvement of the recommendation system. This conversion process enables the recommendation system to better understand user needs, improving the accuracy and efficiency of the recommendation. The subscripts i and j in the method can emphasize the relationship between the user and the content type. This representation makes the sub-module clearly express the recommendation score of the user for specific content. In subsequent analysis and calculation, the parameter subscripts i and j help to accurately evaluate the interest degree of the user for specific content, improving the accuracy and efficiency of the recommendation, so that the recommendation system can respond to user needs more quickly.

[0028] Please refer to Figures 1 to 4 , the correlation degree sub-module processing process is as follows: Step one: AFNY j，i，new = AFNY j，i，old + α × (CRSQ j，i - TSD); Step two: establish the loop termination condition; Condition one: |AFNY j，i，new - AFNY j，i，old | < 0.00493; Condition two: the number of iterations is 123 times; Wherein: AFNY j，i，new represents the correlation degree of user i and content type j after iteration; AFNY j，i，old represents the correlation degree of user i and content type j before iteration, α represents the learning rate, which is used to control the step size of update, TSD represents the comparison threshold, which is used to determine whether CRSQ j，i reaches the expected level; CRSQ based on user i's recommendation score for content type j j，i The difference operation with the comparison threshold TSD is performed, and the learning rate α is used to control the association degree AFNY between user i and content type j before iteration j，i，old Output the association degree AFNY between user i and content type j after iteration j，i，new , use the association degree AFNY between user i and content type j after iteration j，i，new To replace the association degree AFNY between user i and content type j in the content preference submodule j，i , thereby continuously optimizing CPAQ j，i and CRSQ j，i And through conditions one and two, the purpose of loop termination and convergence is achieved.

[0029] In this embodiment: This iterative form can dynamically adjust the relevance between users and content types according to changes in user behavior and content. This dynamic nature enables the recommendation system to more accurately reflect the user's current interests and preferences, thereby achieving more personalized recommendations. Through continuous iteration and optimization of AFNY j，i , the recommendation system can gradually reduce the deviation between the recommendation and the user's actual interests, thereby improving the accuracy of the recommendation. This improvement in accuracy helps to enhance user experience and satisfaction. In the big data environment, this iterative form needs to process and analyze a large amount of user behavior data and content feature data, which requires the software information processing system to have stronger data processing capabilities, including data cleaning, preprocessing, storage and analysis, etc. The iterative form reflects the complexity and dynamics of software information when processing and analyzing large-scale data. Through continuous iteration and optimization of model parameters, the software can more accurately capture and reflect the information and patterns in the data, thereby providing users with better services. Compared with the traditional recommendation system based on big data software information processing, this iterative form pays more attention to user behavior. In order to take into account the dynamism and diversity of content features, it no longer relies solely on static user portraits and content features, but optimizes the recommendation effect through continuous learning and iteration. This iterative learning method makes the recommendation system more intelligent and flexible. In general, this iterative form can make more accurate recommendations, so that users can find the content they are interested in more easily, thereby improving user experience and satisfaction. Accurate recommendations help to distribute content to interested users more efficiently, thereby increasing the exposure and click-through rate of content. A better recommendation system can help content creators attract more attention, thereby promoting their creation and development. Through intelligent recommendations, unnecessary waste of resources can be reduced, such as avoiding recommending content to users who are not interested, thereby optimizing resource utilization efficiency.

[0030] In the specific implementation process, the system is composed of multiple sub-modules in this method. The number of likes of user i on content type j is LKA.j，i , the number of comments of user i on content type j CON j，i , the number of shares of user i on content type j SAE j，i , and the activity of user i ACCTY i Input into the user behavior submodule, output the behavior feature value of user i on content type j UPSA j，i This submodule can comprehensively consider multiple behavior features of the user, such as the number of views, the number of clicks, the dwell time, the number of shares, and the comment content, and thus more comprehensively reflect the interests and preferences of the user. This multi-dimensional analysis makes the user portrait more stereoscopic and accurate, and UPSA j，i provides basic data for subsequent personalized recommendation. By deeply analyzing the behavior features of the user, this submodule can identify the unique interests and needs of the user, and thus provides strong support for personalized recommendation. UPSA j，i not only considers the basic attributes of the user but also deeply analyzes the behavior features of the user, making the user portrait more comprehensive and accurate. This submodule can also dynamically reflect changes in the interests of the user, improving the timeliness and accuracy of the recommendation. The behavior feature value of user i on content type j UPSA j，i is input into the content preference submodule, and based on the popularity of content type j PS j and the association degree of user i and content type j AFNY j，i , the extended preference prediction value of user i on content type j CPAQ j，i is output. This submodule can comprehensively consider multiple factors such as the behavior features of the user, the popularity of the content, and the association degree of the user and the content type, and thus more accurately predict the preferences of the user for specific content. This submodule can recommend content that is more in line with the interests and needs of the user, and thus improve the diversity of the recommendation. j，i The calculation can provide the basis for the optimization strategy of the recommendation system. By analyzing the user preference prediction results, the recommendation system can adjust the recommendation strategy to improve the accuracy and efficiency of the recommendation. CPAQ j，i The calculation considers more factors, including the behavior features of the user and the popularity of the content, making the recommendation results more comprehensive and accurate. In addition, this submodule can also deeply mine the underlying interests behind the behavior of the user, improving the diversity and innovation of the recommendation. The extended preference prediction value of user i on content type j CPAQ j，i and the behavior feature value of user i on content type j UPSA j，i are input into the recommendation analysis submodule, and based on the timeliness of content type j RCN j , the recommendation score of user i on content type j CRSQ j，i is output. CPAQ j，iThe result predicted based on the user preference can further calculate the recommendation score of the user to the specific content, which can intuitively reflect the interest degree of the user to the content and provide accurate scoring basis for the recommendation system. The sub-module can provide more accurate scoring basis for the recommendation system, so that the recommendation result is more in line with the expectations and needs of the user, CPAQ j，i The calculation can provide basis for optimization and improvement of the recommendation system. The recommendation system can find potential problems and improvement direction through analyzing the distribution of the recommendation score, so as to improve the accuracy of the recommendation. The sub-module considers more factors including user behavior characteristics, content preference prediction, etc., so that the scoring result is more comprehensive and accurate. In addition, the sub-module can also comprehensively consider the multi-dimensional relationship between the user and the content, thereby improving the objectivity and fairness of the scoring. Through the CRSQ j，i The difference operation with the comparison threshold TSD is used to iterate the association degree AFNY of the user i and the content type j. j，i The iteration form can dynamically adjust the association degree of the user and the content type according to the user behavior and the content change. The dynamic nature enables the recommendation system to more accurately reflect the current interest and preference of the user, thereby realizing more personalized recommendation. Through continuous iteration and optimization of AFNY j，i The recommendation system can gradually reduce the deviation between the recommendation and the actual interest of the user, thereby improving the accuracy of the recommendation. The iteration form reflects the complexity and dynamic nature of software information in processing and analyzing large-scale data. Through continuous iteration and optimization of the model parameters, the software can more accurately capture and reflect the information and patterns in the data, thereby providing better services for the user. Furthermore, the overall multiple sub-modules can be calculated in cooperation with each other, and the overall circulation and iteration can also be performed, so that the overall system has the effects of automatic optimization and updating, and the adaptability is better.

[0031] Embodiment two: please refer to Figure 1 , Figure 2 and Figure 3 The acquisition module captures user behavior data through web crawler technology, obtains user data through log analysis software, and obtains user behavior data through API interface calling tool. The data cleaning in the data processing module is to remove duplicate data and correct error data. Data integration is to integrate data from different sources and standardize the data to ensure the consistency of the data format. Normalization processing is to convert the data into a format suitable for processing, and then normalize the data to ensure that the data is on the same scale.

[0032] In the embodiment, the network crawler software specifically adopts BeautifulSoup, the API interface calling tool specifically is Postman, and the user's behavior and related data can be collected through the network crawler technology and the API interface calling tool, so as to play the concept of software information processing based on big data, and the collected data is reprocessed based on the data processing module, so that the data can be smoothly input into the subsequent module.

[0033] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A software information processing method based on big data, characterized by: This includes the following information processing methods: Step I: Collect behavioral logs from social media, video websites, and news applications through the collection module to obtain the number of users' likes, comments, shares, and the user's online time; Step II: The data processing module cleans, integrates, and normalizes the number of likes, comments, shares, and online time of users. The data processing module outputs the number of likes by user i on content type j, the number of comments by user i on content type j, the number of shares by user i on content type j, and the activity level of user i. Step III: Input the number of likes by user i for content type j, the number of comments by user i on content type j, the number of shares by user i on content type j, and the activity level of user i into the information processing module. The information processing module outputs the behavioral feature value of user i for content type j, the predicted value of user i's extended preference for content type j, and the recommendation score of user i for content type j. Step IV: Analyze the behavioral feature values ​​of user i for content type j, the predicted value of user i's extended preference for content type j, and the recommendation score of user i for content type j through the analysis module to gain a deeper understanding of user needs and preferences and recommend content that best suits their preferences.

2. The software information processing method based on big data according to claim 1, characterized in that: The information processing module includes: a user behavior submodule, a content preference submodule, a recommendation analysis submodule and a relevance submodule.

3. The software information processing method based on big data according to claim 2, characterized in that: The calculation formula of the user behavior submodule is as follows: UPSA j，i =(LKA j，i ×SA+CON j，i ×SB+SAE j，i ×SC) 2 +ACCTY i -QLT j -0.5 ; in: UPSA j，i represents the behavioral feature value of user i towards content type j; j represents the content type, which is the different types of content in the field of social network analysis. The content types are text, pictures, videos and audio, LKA j，i represents the number of likes of user i on content type j, CON j，i represents the number of comments made by user i on content type j, SAE j，i表示 Number of times user i shares content type j, ACCTY i Indicates the activity of user i, QLT j represents the quality of content type j, SA represents the weight coefficient of likes, SB represents the weight coefficient of comments, and SC represents the weight coefficient of shares; The processing process of the user behavior submodule is as follows: the number of likes LKA of user i for content type j j，i , the number of comments by user i on content type j CON j，i , the number of times user i shares content type j SAE j，i and user i's activity ACCTY i Input to the user behavior submodule, output the behavior feature value UPSA of user i for content type j j，i .

4. The software information processing method based on big data according to claim 3, characterized in that: The calculation formula of the content preference submodule is as follows: CPAQ j，i =(UPSA j，i ×PS j )+(AFNY j，i ×FRHE 0.5 )-(DTET j，i / PLAT i ); in: CPAQ j，i represents the predicted value of user i’s extended preference for content type j; PS j represents the popularity of content type j, AFNY j，i represents the association between user i and content type j, FRHE represents the freshness of content type j, DTET j，i Indicates the degree of user i’s disinterest in content type j, PLAT i represents the popularity of user i, UPSA j，i ×PS j represents the weighted effect of user behavior characteristics on popular content; The processing process of the content preference submodule is as follows: the behavior feature value UPSA of user i for content type j is converted into j，i Input to the content preference submodule and based on the popularity PS of content type j j and the association degree AFNY between user i and content type j j，i Output user i's extended preference prediction value CPAQ for content type j j，i .

5. The software information processing method based on big data according to claim 4, characterized in that: The calculation formula of the recommendation analysis submodule is as follows: CRSQ j，i =(CPAQ j，i ×RCN j )+(UPSA j，i ×TA) + (MTC 0.5 ×PPS); in: CRSQ j，i represents the recommendation score of user i for content type j; RCN j Indicates the timeliness of content type j, TA indicates UPSA j，i The weight factor of MTC represents the matching degree between user i and content type j, and PPS represents the weight factor of MTC. The processing process of the recommendation analysis submodule is as follows: the extended preference prediction value CPAQ of user i for content type j is converted into j，i and user i’s behavioral feature value UPSA for content type j j，i Input to the recommendation analysis submodule and based on the timeliness RCN of content type j j Output user i's recommendation score CRSQ for content type j j，i .

6. The software information processing method based on big data according to claim 5, characterized in that: The processing process of the relevance submodule is as follows: Step 1: AFNY j，i，new =AFNY j，i，old +α×(CRSQ j，i -TSD); Step 2: Establish the loop termination condition; Condition 1: |AFNY j，i，new -AFNY j，i，old |<0.00493; Condition 2: The number of iterations is 123; in: AFNY j，i，new represents the association between user i and content type j after iteration; AFNY j，i，old represents the association between user i and content type j before iteration, α represents the learning rate, which is used to control the step size of the update, and TSD represents the comparison threshold, which is used to determine CRSQ j，i whether it has met the expected level; CRSQ based on user i's recommendation score for content type j j，i The difference operation with the comparison threshold TSD is performed, and the learning rate α is used to control the association degree AFNY between user i and content type j before iteration j，i，old Output the association degree AFNY between user i and content type j after iteration j，i，new , use the association degree AFNY between user i and content type j after iteration j，i，new To replace the association degree AFNY between user i and content type j in the content preference submodule j，i , thereby continuously optimizing CPAQ j，i and CRSQ j，i And through conditions one and two, the purpose of loop termination and convergence is achieved.

7. The software information processing method based on big data according to claim 1, characterized in that: The acquisition module captures user behavior data through web crawler technology, obtains user data through log analysis software, and obtains user behavior data through API interface calling tools.

8. The software information processing method based on big data according to claim 1, characterized in that: The data cleaning in the data processing module is to remove duplicate data and correct erroneous data. The data integration is to integrate data from different sources and standardize the data to ensure the consistency of the data format. The normalization processing is to convert the data into a format suitable for processing and then normalize the data to ensure that the data is on the same scale.

Images