Game Props Recommendation Algorithm Based On User Behavior

With the popularity of the Internet and the rapid development of information technology, we have entered the era of information overload, the game industry data generated by the same level of growth. In order to meet the different needs of users, the game provides a rapid increase in the amount of props information, users often lost in a large number of props information, cannot accurately quickly find their own needs. In this case, the demand for the system is more urgent. Recommender system plays the role of the salesman, by recommending props to the user, so that the user can successfully buy props to their own needs. Recommendation system has good development and application prospects, and the future will attract more and more attention.Collaborative filtering is one of the most widely used techniques for recommender systems, which includes collaborative filtering based on projects and collaborative filtering based on users. And traditional content filtering based on direct analysis of content to recommend different, collaborative filtering analysis of user interest, in users find user specified similar (user interest), the similar user evaluation of a project, it is recommended to generate the results of the user specified for this purpose preference prediction. However, collaborative filtering still faces some challenges:when faced with the very large scale data, collaborative filtering algorithm is faced with the problem of recommendation accuracy and scalability.In view of the above problems, this paper proposes an improved collaborative filtering algorithm based on user behavior of the game props recommendation algorithm. Firstly, we use factor analysis to reduce the dimension of the original data, and get some non-related composite properties. According to the dimension reduction of user attributes, the use of K-means clustering algorithm for users to classify, get a collection of similar users K. Then, using the optimized project based collaborative filtering technology to generate the recommended results. Collaborative filtering module is divided into two steps:the first step is to extract users to buy items of the same class, so as to provide further classification; the second step, in the first step the user set based on Jaccard distance formula of similarity calculation of the costumes and props. Finally, the traditional collaborative filtering algorithm based on the project as a comparative experiment, and step by step to verify the user behavior based game props recommend algorithm can improve the recommendation accuracy.In this algorithm, the recommendation to the user calculation is defined on the set of users in similar, effectively solves the algorithm scalability problem; due to the similar users sense of interest projects usually have a similar feature, therefore improving the recommendation accuracy.