Research And Application Of Personalized Recommendation Algorithm Based On Slope One

Now that the Internet is booming,a lot of products are available for users to choose from.But at the same time,the problem of information overload on major platforms is becoming increasingly severe,which makes online users unable to obtain their favorite products quickly and effectively.The recommendation system can solve this problem.The recommendation system can analyze the information of a large number of users,discover the interests of users,and recommend products for users actively,so that users can get more intelligent and humanized experience and services.Due to the huge practical application requirements of the recommendation system,various related algorithms are constantly being introduced,and recommendation algorithms have become a research hotspot for researchers at home and abroad.Slope One,as a classic kind of collaborative filtering recommendation algorithm,is very commonly used in various websites,but there is also a lot of room for optimization.Therefore,we combines the existing research and takes into account the common problems of the current recommendation system,and optimizes the Slope One algorithm in terms of its sparsity,scalability,Matthew effect,and time factors.The main research contents include:(1)Preprocessing the original rating matrix multiplied by the product popularity weight to alleviate the Matthew effect that often occurs in the website recommendation process;(2)Because the traditional Slope One algorithm does not take into account User similarity and product similarity,this article incorporates user similarity and product similarity suitable for sparse data into the weighted Slope One algorithm;(3)Considering the factors that user interest will change over time,the time weight is incorporated into the weighted Slope One algorithm prediction Score;(4)In order to improve the scalability of the algorithm,this paper introduces the K-Means Plus clustering algorithm to cluster users,and categorizing similar users can greatly reduce the amount of calculation for searching similar users in their neighbors.Finally,the data set of the famous e-commerce platform Amazon is used to verify the performance of the improved algorithm.Experiments have proved that the improved algorithm has higher accuracy in predicting customer ratings and better recommendation effects than the original Slope One algorithm and the weighted Slope One algorithm.