Research On E-commerce Recommendation System Based On Spark Platform

In today's society,the Internet industry is developing rapidly.How to allow users to obtain valuable information from e-commerce websites has become a problem for ecommerce practitioners.E-commerce recommendation system has become the main means to solve the problem.At present,the products recommended by some ecommerce recommendation systems for users do not match the user's interests,and low recommendation accuracy has become a major problem faced by e-commerce recommendation systems.Not only that,when a new user uses the system,the system does not have the user's behavior data,and the system does not know how to recommend to the new user.This is the cold start problem of the system.The recommendation efficiency and the operating efficiency of the system also affect the user experience.Therefore,this article mainly conducts the following research on the shortcomings of e-commerce recommendation systems:(1)Aiming at the problem of low recommendation efficiency,this paper improves the traditional K-Means algorithm,and proposes a K-Means clustering algorithm based on isolated forest—IFK-Means algorithm.The method of clustering the customer's customer satisfaction degree is adopted to reduce the calculation amount of the recommendation algorithm.The IFK-Means algorithm calculates the abnormal factor of the data through the isolation forest algorithm,and screens out the data whose abnormal factor is higher than the threshold.Then in the K-Means clustering process,the Euclidean distance formula and the average error formula are weighted by the abnormal factor.The IFK-Means algorithm improves the clustering effect and the operating efficiency of the recommendation algorithm.(2)Aiming at the problem of low recommendation accuracy,this paper improves the user-based collaborative filtering algorithm,and proposes a collaborative filtering algorithm that integrates customer satisfaction satisfaction—CS-CF algorithm.Analyze the process of calculating user similarity,and integrate user customer service satisfaction into the process.The CS-CF algorithm uses IFK-Means to cluster users' customer service satisfaction.Calculate the customer service satisfaction similarity between users and users through the Euclidean distance formula,and use the customer satisfaction similarity as the weight and the weighted sum of the Pearson similarity to obtain the final user similarity.Then according to the user similarity,predict the user's rating of other products.This method improves the recommendation accuracy of the algorithm.(3)Aiming at the cold start problem,a method of recommending recent hot products is adopted.The system will recommend recent hot products calculated by offline recommendation to new users,alleviating the user's cold start problem.The architecture of the system mainly includes five parts: user visualization,integrated business services,database services,offline recommendation services,and real-time recommendation services.In order to improve the operating efficiency of the system,the e-commerce recommendation system in this article uses the Spark highperformance computing framework to reduce the time required for calculation.The experimental results show that the IFK-Means algorithm implemented on the Spark platform improves the effect and efficiency of clustering and is better applied to the CS-CF algorithm.IFK-Means generates better clustering results when clustering users' customer satisfaction.The experiment verifies the CS-CF algorithm through precision,recall,F1 value,and coverage.Compared with the traditional collaborative filtering algorithm,the CS-CF algorithm improves the recommendation accuracy and recommendation efficiency.After testing the e-commerce recommendation system,its functions can be realized,and the user's cold start problem is alleviated.This article has 45 figures,14 tables,and 87 references.