Collaborative Filtering Recommendation Algorithm Based On User Behavior Model And Ant Colony Clustering

With the rapid growth of Internet and information technology, the information and choices that people faces are much more than ever. The situation of short of information has gradually shifted to information overload. It is a great challenge for both the consumers and providers of information. For consumers, it is hard to locate the data needed efficiently; for providers, it is also a problem to make their information outstanding for users to choose. A good Recommendation System can solve the problems by facilitating the consumers find the data they want easily and making data display to appropriate users automatically at the same time, to achieve a win-win situation.Recommendation Systems are important for large online video websites. Usually, the systems establish personalized model of each individual according to the browsing history of the target user and other related users, and then recommend some videos or movies to them. For example, if a user watched a movie named Life of Pi directed by director Li An, we can guess that the user might be likely to watch some other movies directed by the same director. We can then recommend movies to the user accordingly.Collaborative Filtering Recommendation (CFR) algorithm is one of the most widely used and important technologies in online video websites. The measurement of users’ similarity is the core element in the whole algorithm and affects the results of recommendation largely. There are some major problems and difficulties in traditional recommendation technologies as the following:(1) When clustering users, different algorithms have different performance according to the situation. When the matrix of users’ ratings is very sparse, K-means algorithm is difficult to converge and has low accuracy. Ant colony clustering algorithm is a more suitable method and can promote the results to some extent.(2) Most recommendation systems are based on the browsing history of the target user. However, when a new user register in the system, it is hard to generate proper recommendation results. This also happens at the beginning phase of a online website, where there is little user information. It is not possible for the system to take the movie that the target user is watching into consideration.(3) The fields of videos or movies themselves, such as casts, directors, date, and so on, also have relationship with the recommendation. Traditional algorithms didn’t take these factors into consideration. Focused on the above problems, this paper proposed an Ant colony collaborative filtering algorithm on movie recommendation based on user behavior model.