Personalized Recommendation Based On Visualization

With the development of Internet, the amount of information shows an explosive growth, and this phenomenon will definitely lead to information overload, people hope to find information which they interest from huge amount of information. To a certain degree, information retrieval system help people to search information which they interest, but the traditional information retrieval system fails to take the personalization into account which makes the result of search is always same and the amount of search is large. Therefore, how to help people to find information which they interest from amount of information based their characteristics have become to a unsolved problem. Personalized recommendation is proposed for this problem.Most collaborative filtering (CF) mainly has focused on doing experiments on single dataset or datasets with the same characteristics. We present an analysis of 5 CF algorithms: user-based, item-based, item average, item user average, and Slope one. We apply these algorithms for different types of datasets to get a result: which algorithm can get good result for a give type of dataset.The process of personalized recommendation algorithm is usually automated, So the users can't participate in the process of recommendation to result in bad results. Therefore, the human-computer interaction is very important for recommending. Because the traditional force-directed algorithm cannot represent the weighted graph, we present weighted force-directed algorithm which can represent the strength of nodes to apply the process the recommendation.A significant algorithm is proposed which can display the hidden pattern among the graph currently. Next, we apply k-means algorithm to graph to help users find their community. Finally, we present several interactive methods to help users to find their interest. In a word, it can effectively reduce user's cognitive burden and improve the visual effect of recommendation.