Research On Collaborative Filtering Algorithm Based On Fuzzy Theory And Information Entropy

With the development of information technology,the amount of data generated by institutions and individual users has increased dramatically,making it difficult for web users to efficiently obtain valuable information.The recommendation system actively predicts user needs and recommends data that may be of interest to users,which is an effective method to improve the problem of information overload.Collaborative filtering is the most widely used recommendation technology,but due to the defects of interactive behavior habits and fuzziness in data,there are still some challenging problems to be solved in the research of collaborative filtering algorithms.This thesis intends to carry out related research on the fuzziness,sparseness and cold start of the collaborative filtering algorithm.The main research contents include:1.Aiming at the problem of fuzziness in rating and satisfaction evaluation,a collaborative filtering algorithm based on trapezoidal fuzzy similarity is proposed.The algorithm uses the trapezoidal fuzzy number to describe the mapping relationship between the score and satisfaction,the quantity of information is used to improve the fuzzy number similarity calculation method,the fuzzy scoring prediction strategy estimates the target score to improve the recommendation performance of the collaborative filtering algorithm.The experimental results show that the algorithm can more accurately describe the mapping relationship between score and satisfaction,alleviate the fuzziness of the rating and improve the accuracy of the recommendation.2.Aiming at the deficiency of the traditional collaborative filtering algorithm ignoring the integral features of scoring,a collaborative filtering algorithm based on fuzzy weights and information entropy is proposed.The algorithm uses information entropy and fuzzy score deviation values to describe the distribution information of scores,it uses score deviation fuzzy vectors to express user satisfaction,it evaluates the controversy of items based on score information entropy,and finally uses improved Pearson similarity measure to calculate user similarity.Experimental results show that the algorithm has high accuracy and obvious performance improvement when faced with sparse data.3.Aiming at the sparsity and the fuzziness of the item's tag membership,a collaborative filtering algorithm based on tags and fuzzy similarity was proposed.The algorithm expands the membership of tags to items from {0,1} to [0,1] to improve the similarity of items based on tags,and at the same time,the algorithm use the similarity based on fuzzy scores to form an item similarity measure for collaborative filtering recommend.Experimental results show that the algorithm can suppress the problems caused by the cold start of the project and the sparsity of the score data to a certain extent.