Design And Implementation Of Weibo Friend Recommendation System Based On SSD And Timing Model

With the continuous development of the Internet,making friends and chatting on social media has gradually become a part of people's daily life.Users create a large amount of data every day,crawling and analyzing these data is a common business behavior on various social media.On social media,because the merchants can not effectively and reasonably make use of the complex data created by users,so how to analyze users' information and optimize the platform's friend recommendation strategy is the main content of the research.Through the research of the group of microblog's users,the paper finds that the recommended factors considered by users are often hidden in their published microblog.Pictures can convey information no less than words,and hobbies will fade over time.Combined with the above characteristics,by using the target detection technology to extract the information in the pictures,disposing the extracted results with the time series model,I have designed a microblog friend recommendation system.The main work of the paper is as follows:(1)The acquisition of microblog users' data set: based on Sina microblog,this paper completes the implementation of crawler system based on Scrapy framework.The system adopts simulated login method to solve the problem of identity authentication of microblog users,using proxy pool,cookies pool to solve the problem of anti-creptile,building the URLthrough the analysis of the Ajax links of each page in microblog,realizing the recursive crawl and the collection of users' personal data,friend information,microblog data based on the Scrapy framework(2)The proposition of friend recommendation algorithm based on SSD and Time Series model: using the collected information of users such as gender,age to construct the friend recommendation which is based on demography and to obtain the similarity between users.According to extracting the pictures in the collected microblog information and multi-objective detection algorithm SSD to obtain the result of interest classification.According to extracting the user's publication time to set the time interval and to dispose the results of interest classification combined with the time series model,also to obtain the score of each interest.Then the nearest neighbor set is obtained by calculating the similarity according to the cooperative filtering idea.Finally,the above two modules are linearly fused to obtain the final similarity data.And this article uses the TOP-K idea to recommend the friend.Based on the real data set of Sina micrblog,the accuracy,recall and F values were calculated and verified by experiments.It verifies the validity and accuracy of the proposed friend recommendation algorithm.(3)The design and implementation of the friend recommendation system based on Sina microblog: the front end uses the Vue.js to develop the system page,the back end uses the microblog crawler,the MySQL database and so on to realize the dispose of data,and applies the friend recommendation algorithm based on the SSD and timing model to the system,so that the recommendation result can be displayed better and easy for the user to view.