Hybrid E-commerce Recommendation System Based On Big Data Platfor

With the gradual maturity of the Internet and information technology,the massive data generated by various mobile devices has brought people into the era of information overload.It is increasingly difficult to mine information that is conducive to human development and social progress from the massive information.Online shopping is concerned by more and more consumers and gradually becomes an irreplaceable consumption mode.However,it has become a big problem for many consumers to choose their favorite products from a variety of products.Recommendation system is an important tool to solve this contradiction.Although In the nearly30 years of development,the cold start problem,timeliness problem,and the optimization of model super parameters have always affected the recommendation performance of the recommendation system.(1)When a new user logs in to the system for the first time,the system cannot obtain the user’s detailed information and cannot make recommendations to the user,resulting in a system cold start problem.In view of this situation,the average score of each product is calculated periodically as a popular recommendation for high-quality products according to the ratings of all users on the products in the historical data.(2)As time goes on,users’ interests and hobbies are also changing accordingly.Recommendations made for past users’ interests and hobbies must not be favored by users.The system uses big data ecosystem components to form a log collection system to collect user buried point data in real time and make real-time recommendations based on similar recommendation algorithms.(3)There is a large amount of sparse data in the user product rating table.This paper uses matrix decomposition method to solve the problem of data sparsity when designing the offline recommendation system.The quality of the matrix grading model is affected by the super parameters.Therefore,the gradient descent algorithm commonly used in machine learning is used to optimize the model super parameters.Based on the above work content,this paper designs a hybrid recommendation system.The improved dynamic hybrid recommendation algorithm is used as offline recommendation,and the big data ecosystem component is used as the log collection system to collect user buried point data in real time to generate an online recommendation system.