Design And Implementation Of Cross-Border E-Commerce Management System Based On Recommendation Algorithm

In order to meet the needs of the cross-border e-commerce market,many cross-border e-commerce companies have emerged.Each cross-border e-commerce company offers similar products,and the competition between different cross-border e-commerce companies lies in service quality and more accurate product recommendations,accurately reaching the corresponding consumers.Because the types of cross-border e-commerce products are often very diverse,and there are significant differences in the goods that each customer needs to purchase during the consumption process,how to use big data and information tools to recommend the target products they want to purchase to customers within the scope of cross-border e-commerce products that the company can supply at the fastest speed has become a very practical issue.For cross-border e-commerce companies,innovating service ideas and methods and adopting information technology to establish a management system with product recommendation requirements to meet the common business needs of cross-border e-commerce companies is an important means to improve the service quality of cross-border e-commerce companies.This thesis is based on real and effective cross-border e-commerce transaction business data,and combines mainstream trend based product recommendation technology and algorithms to achieve the following research work:1.On this basis,the preprocessing and feature extraction of cross-border e-commerce related business information were completed.Pre processing and feature extraction of data stored in the SQL Server database for actual cross-border e-commerce business,such as logistics,goods,customs declaration,and orders,to provide data support for the product recommendation technology used in this paper.2.Design,develop and implement a cross-border e-commerce management system based on actual business scenarios.The system is a Java Web project based on B/S mode.The front-end information transmission is requested through Ajax,and the back-end adopts the popular MVC design mode.The login module uses the Apache Shiro security manager to execute authentication and authorization programs and Shiro framework,while using the homepage data of Shiro tags to achieve information confidentiality.3.Adopting a Siamese network model based on deep learning,the twin tower model achieves product recommendation function.It involves the minimum maximum normalization method,Re LU activation function,mean square error(MSE)loss function,Batch Stochastic Gradient Descent(BSGD),and cosine similarity calculation method.By constructing corresponding portraits of users and goods and according to some characteristics of users,it can recommend goods purchased by similar users or similar goods.The main innovation of the paper lies in adding specific algorithms to the traditional classic recommendation methods to match them with real business scenarios,and improving them to use actual customs clearance data as experimental data for the recommendation algorithm.By integrating deep learning based Siamese network models and twin tower models,the recommendation results are more realistic.