Design And Implementation Of An Online Product Recommendation System Based On Stream Computing

Due to the continuous maturity and rapid development of network information technology,a new type of data-intensive application has emerged in the fields of web applications,network monitoring,sensor monitoring,telecommunications,finance,manufacturing,etc.It's called streaming data,that is,data in large amounts,Fast,time-varying flow patterns continue to arrive.How to collect and calculate these stream data to generate actual economic value is a hot topic of network technology research in recent years,and there is a huge amount of product information in e-commerce websites.Users find a product that suits them is no different from a haystack So,develop an online product recommendation system.This recommendation system can make use of the streaming data generated when users browse and purchase products,discover the users' immediate purchase intentions and interests,and find products similar to their interests for users,and recommend them to users.It can not only improve the sales of goods,but also increase the satisfaction of customers.The online product recommendation system based on stream computing includes a data collection module,an online recommendation module,a solution to the cold start problem,an offline calculation module,and a web display module.The data collection module is implemented using the Flume framework.Its main responsibility is to collect the user's online log data information(browsing,collection and purchase),and then transmit it to the distributed message queue Kafka.The off-line computing module is to analyze and calculate the past behavior information of users.It uses the idea of collaborative filtering recommendation model based on items and the idea of recommendation model based on tags,the Flink flow computing framework is used to calculate the list of product recommendations,and stores the calculation results in the database;the main function of the online recommendation module is when the user has an operation,Discover users' immediate purchase intentions and interests online,and use the calculation results of the offline calculation module to recommend related products to users in real time;the Web display module provides some external interfaces for offline recommendation results,popular products and online recommendation results A concise display.The results of testing the system show that the system can discover the user's immediate purchase intentions and interests online,recommend products similar to their interests to users,and can run stably for a long time to achieve the design goals.Therefore,the recommendation system in this article can be used in the field of business,we recommend customized products related to their interests for users.