A Collaborative Filtering Recommendation Algorithm Research Combining Occupancy And Time Decay

The development of information technology and the rapidly increasing number of Internet users have brought about the problem of information overload. Collaborative filtering recommendation is an important method to solve this problem. However, using collaborative filtering personalized recommendation applications have been plagued by sparse data, scalability problems and other serious problems. The above problems are effectively alleviated by introduction of a new factor from different angles. In the background film recommendation, this paper contains the following aspects.Firstly, this paper analyzes the advantages and disadvantages of traditional collaborative filtering. Taking into account the actual situation of user interest happens to change with the passage of time, a time decay factor is added into similarity calculation step in the traditional collaborative filtering algorithm, to reduce the adverse effects of time factor on the recommendation result.Secondly, The traditional similarity measure method will get a similarity which is either too high or too low on the condition of data sparsity. Considering this situation, occupancy degree weight coefficient is put forward to optimize similarity calculation, which makes recommendation neighbor based on the similarity of the results more reliable to improve recommendation accuracy.Thirdly, the concept of frequent degree is proposed integrating the advantages of item-based and user-based collaborative filtering method. We dynamically strategy integrate the item-based and user-based on collaborative filtering method through the degree of frequent support factor, then a hybrid algorithm based on a time decay is used for the hybrid algorithm recommendation.Finally, experiments are designed on the foundation of all the above ideas, to implement the hybrid algorithm combining occupancy and time decay. The standard of mean absolute error is accepted to evaluate the experimental results and to check the validity of the occupancy weighted, time damped and frequency sustained mixed algorithm. Experimental results show the effectiveness of the approach in improving recommendation accuracy and alleviate the problem of sparsity.