Research On Personalized Recommendation Algorithm Based On Covariance

With the development of the Internet and e-commerce,people can choose and buy their favorite items online without leaving home,and the major e-commerce companies in order to be able to improve revenue or improve the quality of service,one after another put forward and improve their own personalized recommendation system.Up to now,many recommended algorithms have been put forward,the recommendation system needs to ensure that the accuracy of the recommendation and recommendation diversity and novelty to achieve a good balance,which is an important direction of the current recommendation system algorithm research.On the basis of covariance theory,this paper regards the user's selection of an item as a random event,and the recorded value generated between the user and the item can be regarded as a random variable,and the probability of the user selects any item is regarded as an equal probability distribution.This assumption is exactly full of the characteristics of the covariance theory to study the relationship between samples without knowing the distribution of random samples in probability statistics.Therefore,based on covariance theory,this paper forms covariance matrix based on covariance values to quantify the relationship between items,and constructs a recommendation model based on covariance.The work of this paper is as follows:1.Building recommendation models based on covariance: analyze the relationship between covariance theory and experimental data,and improve the algorithm based on covariance recommendation model;The personalized recommendation algorithm based on covariance proposed in this paper can achieve the accuracy of recommendation and improve the diversity and novelty of recommendation results.2.The experimental scheme is designed: including the preprocessing of the public data set,the test division of the data set and the implementation of seven commonly used indexes to check the performance of the recommendation model and the design of experimental group.3.The performance analysis of algorithms: This paper analyzes the reason why the method proposed can improve the diversity and novelty of recommendation results while achieving the accuracy of recommendation.In addition,a comprehensive analysis of the covariance-based recommendation algorithm and its improved algorithm.The personalized recommendation algorithm based on covariance and the improved algorithm based on covariance in this paper is parameter-free.Compared with the existing algorithms on the published data and experiments,the experimental results show that the covariance-based algorithm and its improved algorithm proposed in this paper have good performance in accuracy,diversity and novelty.Therefore,the research provides a new research method for existing recommendation algorithms,which is of great significance to the application of the recommendation algorithm.