Collaborative Filtering Recommendation Algorithm Based On Timeliness Modeling And Feature Fusion

With the rapid development of the Internet,information overload is a common phenomenon in present era.Facing massive information,how to extract the content that users interested in becomes a key issue for Internet services.Collaborative filtering recommended technology mining the user's historical behavior,modeling the user's interest,and predicting the user's potential behavior,has been widely used in a variety of Internet services.However,the current collaborative filtering algorithms can not reflect the changes of the user's interest in time,and the recommendation accuracy is poor when the user-item scoring matrix is sparse.In this paper,we focus on the following two issues,modeling the changes of the user's interest and item similarity respectively,and researching collaborative filtering technology based on time modeling and feature fusion.The main work includes:A collaborative filtering recommendation algorithm based on the Ebbinghaus Forgetting Model is proposed to solve the changing of people's interest with the passage of time.The people's interest is changing over time,which is a natural forgotten process that can be described by forgotten curve.The algorithm uses the forgetting function to model the change of human interest,and the time quantization function of the user's score is taken as the time factor to attenuate the user ' s original score,in order to reflect the changing of the user's interest.The time factor of the score as the weight of the rating attribute is introduced into the calculation of user similarity,which can effectively improve the accuracy of recommendation.In addition,taking into account the ratio of the number of two users to the total score and the introduction of the contraction factor to calculate the similarity correction,to build a calculation model based on Ebbinghaus forgotten similarity,and then calculate the user's neighbor set according to the model and finally use the score forecast formula to select the recommended items to the target user.The experimental results show that the improved method can improve the prediction accuracy effectively.A collaborative filtering recommendation algorithm based on feature fusion is proposed to solve the problem of poor accuracy in sparse user-item scoring matrix.The algorithm mines the item similarity from both item attributes and item scores,and constructs a comprehensive similarity measure formula with weight to predict the unknown score in the score matrix.And then according to the user similarity calculation model to find the target user's neighbor set,and then according to the scoring forecasting model to predict the target user's interest in the unrated items and Identify the items to be recommended.The algorithm increases the data density of the user-item scoring matrix and alleviates the inaccurate problem of similarity calculation caused by the user-item scoring matrix sparsity.The experimental results show that the method can improve the prediction accuracy effectively.