Social Network Recommendation System Based On User Reading Interest: Design And Implementation

In recent years, internet Social Platform has already developed and expanded extremely, which makes the improving service and performance more and more efficiently. Nevertheless, the internet Social Platform which faces to the article reading area has not really developed. On the traditional reading application software, we can share and exchange but with the help of the third-party social platform. The reading social system that integrates article reading and social function, adds the function of article annotation. The development of the social system can't do without information recommendation technology, and the recommended system needs to solve the problem: how to do well in article classification and the user's reading interests mining with the massive article data and user's data.The paper uses article classification and the user's reading interests mining as the core issue of recommended system. Firstly, it contrastively analyzes the current common text classification technology according to the internet article features. Eventually, it can extract feature words by using the method of multi-feature fusion, and classifies by using supporting vector machine (SVM). And then reach the destination of user's reading interests mining. The paper analyzes user's reading behavior, including collection, annotation, comment, browse and so on, and then determines the user's interest tendency.it uses statistics method to establish user's reading interest model, and updates the user's interest model by forgetting calculation. Finally, by the included Angle cosine similarity algorithm the paper will find out the similarity between different models, and then recommend user the information adopting Content-based Recommendation Algorithm.Tests show that multiple features fusion method to extract key words on the processing of text categorization has much better result than information divergence and chi-square statistic method. Through statistical methods to analyze the user's reading behavior we can effectively express the user's interest in reading. Meanwhile recommendation algorithm based on the content in this system has good accuracy, and the accuracy of the article recommended achieved at least 80%, meets the needs of the project.This paper firstly introduces the background of project, and then presents the problem. To solve this problem, the paper compares different technical way and then chooses the best way to identify the problem solutions. After analyzing and defining the demand of the system, the paper gives a detailed introduction to the recommended scheme, and then analyzes the architecture of the system and the implementation of the core function. At last, the paper verifies the recommended scheme, which proved the feasibility of the recommended scheme.