Research On Personalized Film Recommendation Algorithm Based On Context Sensing

Recommender System’s purpose is automatically recommend some related items to users, including books, movies, songs, tourist spot, etc. The existing recommendation system uses the technology in general two types: content based filtering and collaborative filtering. Collaborative filtering only consider the user ratings matrix of objects, and less consideration of the situation. However, in practical application scenarios, user interest in articles usually affected by age, occupation, affect the surrounding environment and other factors. At the same time, the popularity of items with the time and the surrounding environment changes. In recommendation system into the context information can effectively improve the recommendation precision. With the popularization and application of e-commerce website and music, movies, travel websites and the rise of mobile terminal services, personalized recommendation is becoming more and more important. The recommendation system contains a variety of context data, such as mo vie recommendation contained time, companion, weather, discounts and other information. The context information of user and item reflects the real environment. Therefore, the context into recommendation algorithm can improve the recommendation performance.This paper first studies the field of recommender system algorithms, and analyzed their advantages and disadvantages, try to integrate all kinds of context information related to the base of the traditional recommendation algorithm and recommendation performance to get better. Existing recommendation algorithm in the prediction of the scores and recommending items, less consideration of consumer goods, the purpose of the context information, so that the recommend’s effect is not good.This paper tries to introduce the contextual information in collaborative filtering algorithm, according to the split of context information for specific items.The ratings of certain items are split, according to the value of an item-dependent Contextual condition. For example the film score, when a film in a specific condition score much higher than other under the condition of the score, the movie will be split.We use the MovieLens data set and some of the semi synthetic data sets algorithm and nearest neighbor collaborative filtering algorithm based on matrix decomposition experiment. The experimental results show that the method, situational split items based on recommendation accuracy index is better than existing algorithms.