Design And Implementation Of Financial Product Recommendation System Based On Zhihu User's Behavior

Financial institutions need to explore and develop new customers for the sale of financial products.According to the browsing habits of users of social networking sites,financial institutions can recommend and develop potential customers.Financial institutions' financial products are mainly located in high-income,highly educated customers,knowing that users have the above characteristics,so this paper chooses the topic of financial products recommendation system based on knowing users' behavior.The main work of this paper includes:(1)Design and implement the crawler.Used to crawl all users' browsing information under the "Finance" section.The crawler is designed and implemented.Firstly,the web page is acquired by get method under request,and the web page is parsed by xpath technology.The acquired data is saved in JSON format with pandas framework.The data obtained in this way are mainly the information of answering questions.At the same time,with the help of the voting function under the "financial management" section of the forum,user data are classified and marked according to the degree of risk they can bear.One kind of data is the data with the marker of user's risk tolerance,and the other is the data without marker.(2)Preprocess the data acquired by the crawler and discover the users who are interested in the financial products under the "financial management" plate.The data acquired by the crawler has many problems,such as too long text,too many useless words,and many uncertainties in user's answers.Therefore,it is necessary to pre-process the crawled user data.Pre-processing mainly includes the use of word segmentation technology for long text segmentation.Using the open source framework of Jieba word segmentation for reference,this paper studies the algorithm of Chinese word segmentation.By designing the stop word list and user-defined word list to filter useless words and symbols,TF-IDF algorithm is used to discover the users who are interested in financial management,and the processed data are the highly relevant text data and text vector of the word segmentation on the topic of "financial management" for recommendation.(3)Naive Bayesian and KMEANS clustering recommendation algorithms are used to implement the recommendation system.This paper studies the current mainstream recommendation algorithm.On the user data that may be interested in "financial management",the recommendation algorithm is used to recommend financial products with different risk levels for users.Among them,Naive Bayesian algorithm and training model are used to calculate the recommendation results for data sources with the marker of user's risk tolerance.The recommendation results of Naive Bayesian algorithm are analyzed by F1 score,accuracy,recall rate and other indicators.For unlabeled data,KMEANS algorithm is used to train the model to achieve data clustering,and the performance of the algorithm is analyzed by contour coefficient.The recommendation results are a list of risk tolerance of users and financial products for use by financial institutions.(4)The financial product recommendation system designed and implemented adopts B/S structure and MVC design mode.The recommended results are connected to private trust pages for the use of financial institution's marketing personnel.