Listed Company Recommendation Service Based On Improved Collaborative Filtering Algorithm

With the improvement of people's quality of life,people gradually realize the importance of financial management.More and more people no longer traditionally put money in banking institutions,but instead buy funds,stocks and other financial products.As of the end of 2017,the size of domestic funds has exceeded 10 trillion yuan,and more than 100 fund management companies have been listed.For investors,it is natural to choose the fund company that makes them the best investment income,and for fund companies,they are faced with the problem of how to provide fund recommendation for people more efficiently.Correctly assessing the investment value of a company is the basis for a fund management company to provide people with reasonable investment decisions.Fund management companies often face the problem of being unable to obtain sufficient corporate index data when conducting corporate value evaluations.There are many solutions to this problem at the moment,and personalized recommendation technology is one of them.Collaborative filtering algorithm is one of the most widely used methods in the recommended technologies.The basic idea of collaborative filtering is that the method is to predict the user's interest according to the user's history rating data so as to achieve personalized recommendation.However,using the traditional collaborative filtering algorithm directly cannot accurately predict the missing indicator value of the company,and thus it cannot help the fund management company to generate an effective recommendation of the listed company.In order to solve the above problems,this paper proposes a method for recommending listed companies based on improved collaborative filtering algorithm.The main research work of this paper is as follows:(1)Aiming at the problem that the traditional collaborative filtering algorithm is not ideal when the data is sparse,this paper proposes a collaborative filtering algorithm based on hidden relations.First,a hidden relationship between users is constructed based on the user's common rating information,and potential relationships between users are discovered through the hidden relationship.Then this hidden relationship is combined with Pearson correlation coefficient to improve the accuracy of Pearson's correlation coefficient in the high-dimensional sparse matrix.(2)For the problem that the traditional collaborative filtering algorithm is not ideal when facing new user problems,this paper proposes a collaborative filtering algorithm based on similarity local filling.First,the project-based collaborative filtering algorithm is used to initially predict the possible scores of new users and to partially populate new users.Then,based on the filled score matrix,the user-based collaborative filtering algorithm is used to implement recommendation.(3)This article crawls the publicly released indicator data of several listed companies on the wind website through a web crawler,and proposes a data logarithmization method to alleviate the extreme value problem.Finally,the improved collaborative filtering algorithm proposed in this paper is used to solve the problem that fund companies can't accurately evaluate the operating conditions of listed companies because of lacking sufficient indicator data,so as to help fund management companies to implement recommendations for listed companies.