Research On Movie Recommendation Algorithm Based On User Characteristics And Time Weight

In the context of big data,it is more difficult for people to quickly obtain information of interest from vast amounts of information.The emergence of the recommendation system not only extracts valuable information for the user,but also makes the information rationalized.At present,the collaborative filtering algorithm is one of the widely used recommendation algorithms.However,the traditional collaborative filtering algorithm ignores the influence on the recommendation results in the user cold start and user interests changes over time.Thereby reduces the accuracy of the recommendation.Based on the analysis of existing research,this paper first adds the user characteristics to acquire user similarity and then further improve the similarity according to the user's trust relationship.At the same time introducing the time factor,a time range is set within the difference between the user's rating time and the current time to find a turning point where the user's interest changes.Therefore,this article focuses on the two aspects of research,related research and improvements are as follows:(1)For the traditional collaborative filtering algorithm does not take into account the user cold start problem,this paper added the user characteristics in the calculation of similarity.It can be considered from two aspects: user attributes and user trust.First,user attributes and user trust are selected and modeled separately.The two are combined to find the final similarity,and a more reliable nearest neighbor is selected based on the similarity.(2)Since the traditional predictive scoring is easily affected by the user's interest transfer,the existing collaborative filtering recommendation does not consider how to mine the time when the user's interest changes while giving the scoring time weight.Therefore,by setting a time range in the time function,this paper finds the law of user interest change according to different values,because the user's interest is not the same in different time ranges.Finally,the time function is integrated into the prediction score calculation formula to obtain a more accurate prediction score.This effectively improves the problem of changes in user interest over time,while also improving the accuracy of recommendations.(3)The experimental data uses the classic MovieLens data set.The experimentalresults show that the improved algorithm in this paper has a smaller average absolute error(MAE)value in the recommendation,and effectively mitigates the cold start of the user and the user's interest over time.The problem,the quality of the recommendation has also been improved.