Research On Automated Operating System Based On WeChat Public Platform

In recent years,The internal mobile Internet has developed rapidly.The influence of the media platform is becoming more and more powerful,and is gradually becoming an important source of information for the public.As an important representative of the media platform,the WeChat public platform has attracted hundreds of millions of users.This thesis will design an automated operation system of WeChat public platform by analyzing the public numbers' issue and WeChat public number related information,and make in-depth research on the two main functions of the system,material classification and article recommendation.The system can provide great convenience for WeChat public operators to find materials.It can reduce operators' workload and publish high-quality tweets,attracting users' continuous attention.The main work of this thesis is as follows:(1)The framework of WeChat public platform automation operation system is designed,and each functional module in the framework is designed and realized one by one.The process of system task execution is expounded,and the functions of data collection,text classification,and personalized recommendation of the article are realized.(2)With the combination of UIautomator and Xposed tools,the WeChat article data collection is completed and stored in the MySQL database,so as to provide necessary data support for the operation of the system.(3)Through the study of the existing feature selection and classification algorithms and further experimental comparison,a feature extraction method suitable for long and short data sets is proposed.And then the text is quantified with GloVe model,finally WeChat article is classified by LinearSVM classification algorithm.The experimental results show that the model is good in the classification of WeChatarticles.The average accuracy reached 87%,the average precision reached 88.7%,the average recall rate reached 86.2%.Compared with the classification without the feature selection algorithm the results improved 1.2%,1.5% and 0.7% respectively.And compared with the traditional classification model,the average F1-Score improved 6.6%.(4)A hybrid recommendation algorithm is proposed in this thesis.And we used the actual user data verified the effectiveness and practicability of the hybrid recommendation algorithm.The results are compared with the results of several other classical recommendation algorithms.The main contributions and innovations of this thesis are as follows:(1)A feature extraction scheme for long and short mixed text is proposed.And we used it implement the WeChat data classification model.And the classification performance of the system is improved.(2)We designed and implemented a personalized recommendation hybrid recommender.The hybrid recommendation algorithm considered the characteristics of users,and is a more suitable algorithm for recommendation of WeChat public number content.It has good expansibility.In the future,we can tap more user characteristics,add new recommender to weighted items,and avoid many shortcomings of single recommendation model,such as new user problem and new project problem,so that each recommender avoids weaknesses.