Research On Recommendation Methods Based On Evolutionary Algorithms

Recommendation system is an effective tool to solve the information overload problem.However,with the massive increase of information,the problems of data sparsity,poor scalability,and cold start of recommendation algorithms have become more and more prominent.In order to improve the recommendation generation process of collaborative filtering methods,some scholars have introduced clustering techniques to group users.The existing literature uses many traditional clustering mechanisms for clustering users,but the use of evolutionary algorithms to optimize clustering techniques for generating optimal recommendations remains to be explored.Also,due to factors such as the design of the recommendation algorithm itself or the characteristics of the data set,the recommendation algorithm also faces the problem of bias that is difficult to handle,which limits the accuracy of the recommendation.To solve the above problems,two aspects of work are done in this paper as follows:We propose a collaborative filtering algorithm based on an evolutionary algorithm to optimize clustering.The algorithm combines the sparrow search algorithm and the clustering algorithm,and uses the sparrow search algorithm to optimize the selection of the centroids of the K-means algorithm,thus improving the sensitivity problem of the centroid selection of the clustering algorithm.This approach generates better clustering results,which improves the clustering effect and increases the scalability of the recommendation algorithm.Also considering that the data in the dataset is observational rather than experimental,a scoring preference model is used to eliminate selection bias and an item type preference model is used to eliminate exposure bias.To address the problems of limited global search capability and premature convergence of the sparrow search algorithm,Sine chaos is used to initialize the sparrow population to improve the quality of the initial sparrow population,and the butterfly optimization algorithm and simulated annealing algorithm are used to improve the update method of sparrow positions to increase the global exploitation capability and avoid premature convergence.In addition to selection bias and exposure bias,recommendation result bias,i.e.,popularity bias,is also considered and a popularity preference model is introduced.Experiments are conducted on the movie rating datasets Movie Lens and Film Trust to compare with other recommendation algorithms,and the experimental results show the superiority of the proposed algorithm in terms of scalability,performance and personalized movie recommendations.