Research And Implementation Of Intelligent E-Commerce Recommendation Platform Based On Collaborative Filtering

With the continuous development of technology and the increasing recognition of e-commerce,online shopping has gradually became an indispensable part of life.Taobao,Jingdong and other platforms have been deeply engaged in the C2 C field.While expanding,they have also changed the traditional business marketing model and promoted the continuous and vigorous development of domestic e-commerce business.The progress of the industry has brought great profits and convenience,but which also faces increasing challenges.First,there is the problem of "data overload".The booming of e-commerce business makes the data of relevant users and commodities grow explosively,and it is more and more difficult for users to find the required information timely and accurately from the complex data.Second,there is the "cold start" problem.There is no corresponding historical behavior data for new users,and the platform cannot recommend suitable products to them through prediction,which is called cold start of user.The same applies to cold start of item.Faced with the urgent problems of "data overload" and "cold start",personalized recommendation system emerges at the right moment.In this paper,the document database mongodb is used.Combined with Alternating Least-Squares based collaborative filtering algorithm,bisecting K-means algorithm,Spark,flume,kafka and other big data technologies,the intelligent e-commerce recommendation platform based on collaborative filtering is designed and implemented.Focusing on cold start and data overload problems,the platform combines a variety of recommendation methods such as offline statistical list,offline recommendation and real-time recommendation.The platform mines suitable commodity groups for users from the numerous and huge data and efficiently obtains valuable information.The main work of this paper is as follows:(1)Through the investigation of the e-commerce industry and the analysis of the recommendation process during the internship,the function and performance requirements of the platform were clarified,and the overall outline design and the detailed analysis and design of sub-modules were made from the perspectives of business architecture and technology level.(2)In view of data overload,this paper used the architecture of big data,Spark,mongodb,flume,kafka and other tools to build an efficient data flow platform,and established a business data warehouse based on star model to store the slowly changing dimensional data and behavioral data of users and commodities in the form of a ziped table.It improved the efficiency of storage and processing of data.(3)Based on the big data platform,this paper constructed a user-product hybrid recommendation model which includes Alternating Least-Squares based collaborative filtering algorithm and bisecting K-means.The hot spot factor is added and the model recommends the mixed list through the Top N algorithm.(4)In order to solve the problem of cold start of user,this paper supplemented the data of new users from the off-line statistical recommendation and the new users' label questionnaire.At the same time,based on the realization of e-commerce business,the intelligent e-commerce recommendation platform based on collaborative filtering is completed by referring to the offline and real-time hybrid recommendation of Lambda architecture.Finally,a test model is constructed to detect the functional completion and performance level of the platform.