Research On Collaborative Filtering Recommendation Algorithm Based On Fuzzy Cognition

With the rapid development of modern technology and the extensive application of computer networks in all walks of life,e-commerce system has been closely related to each of us,so that our lives become more convenient.At the same time,the large and complex information in the e-commerce system will affect our accurate and appropriate choice of goods and services.Therefore,the research of the recommended algorithms comes into being.Among them,the collaborative filtering algorithm has been widespread concern in the recommended community.At present,with the change of users' interests,timeliness problem has become one of the key factors affecting the accuracy of recommended algorithms.This paper analyzes the shortcomings of the existing collaborative filtering recommendation algorithm and timeliness problem of the research works.And then we point out an important problem:the purchase time of the goods is sequential,but the existing collaborative filtering algorithms suggest that ratings produced at different times are weighted equally in different historical period,but this cannot reflect the changes of item ratings,which will result in a decrease in the recommended quality of the recommendation system.In this paper,an improved collaborative filtering is proposed by combining the existing algorithms with the characteristics of timeliness.Experiments on Netflix datasets demonstrate that the effectiveness of our approach is provided to validate.The main work and innovation of this paper is reflected in the following aspects:1.The traditional collaborative filtering algorithms suggest that ratings produced at different times are weighted equally in different historical period,but this cannot reflect the changes of item ratings,which will result in a decrease in the recommended quality of the recommendation system.This paper argues that the ratings and weights are consistent with the law of fuzzy increasing.2.On the basis of the fuzzy increasing of weights,a collaborative filtering algorithm based on time weight is proposed.In the personalized prediction phase of the item,the prediction algorithm is improved by introducing the time function.And in the process of solving the weighting parameters,an improved particle swarm optimization algorithm is proposed.At the same time,it is assumed that the same user has the same interest trend for the items with strong correlation degree.So we constructs the objective function with the individual user and the individual item cluster.In this paper,an item clustering method based on discrete search is proposed by using the rating attribute of the item.3.On the basis of the fuzzy increasing of rating information and weights,the fuzzy cognitive filtering algorithm is proposed.First of all,in order to reflect the trend of item rating in each historical period,this paper proposes a similarity measure based on time window divided by the time information of the item.Then,when discussing the law of item weights in each historical period,we introduce the time function to construct the objective function.Finally,a parametric parameter estimation method is proposed when solving the fuzzy increasing parameters in the objective function.4.The proposed algorithm is valid though designing experiment and analyzing recommended results on Netflix data.Experimental results show that the improved collaborative filtering algorithm has better recommendation than traditional cooperative filtering recommendation algorithm.So it,to some extent,promote the progress of theoretical research and applied research of collaborative filtering algorithm.