The Design And Implementation Of Refund System And Refund Prediction Model For An Eletronic Business Platform

With the popularity of Internet technology and mobile Internet,online shopping online payment has entered the public.Meanwhile,behind a large payment data in the online refund transaction behavior is also rising.Due to the special form of online shopping and liberal refund policy and other factors provided so that the refund rate than offline shopping refund rate compared to high,and traditional refund process involving involving accounts,finance,payment institutions and logistics and other constraints,there is a refund long cycle,low efficiency and other issues.So the introduction of a set of e-commerce online refund system to increase the success rate of online refund,through the analysis of the refund data mining to reduce the refund behavior,thereby reducing the refund rate bring profit to the enterprise is imperative.This paper designs and implements a refund system based on the electricity platform to improve the efficiency of online refund and solve the refun d problem of the multi-payment channels.The refund forecasting model is constructed,From the current refund data to predict whether the new transaction is a refund,the forecast results as a basis for decision-making,thereby reducing the refund rate for the company to increase profits.The refund system uses the SSI framework to adopt an asynchronous refund method that separates the acceptance and refunds,takes efficient and asynchronous refunds based on the JMQ and the Quartz task framework,and adds retry and query mechanisms to increase online Refund success rate.Based on the analysis of the refund data,this paper uses the Borderline-SMOTE algorithm to solve the problem of data imbalance and the stochastic forest algorithm to solve the problem of attribute weight imbalance using the stochastic selection algorithm of feature variables.The experimental results show that the prediction accuracy of the model is 91.6%,and the forecasting result can be used as the decision basis to reduce the refund behavior and reduce the refund rate.