Sales Prediction Based on Machine Learning Approach: Support Vector Machine, Xgboost and Random Forest

Abstract

Living in the age of digitalize, the introduction of the Internet and e-commerce have brought consumers with convenience and choices. Since more people are used to shopping on the Internet, e-commerce enterprises should be able to predict demand of consumers when facing the fierce competition in the rapidly growing market. Therefore, the background of highly competitive and rapidly changing market has proved the significance sales prediction is for an enterprise. An accurate result of sales prediction can support enterprises with business decision basis. Previous studies on machine learning approaches used for sales prediction include techniques such as data mining, deep learning and time series analysis, so as to improve the performance of prediction models. Besides, researchers develop prediction models against specific industries, regions, etc. Through continuous study and innovation, with the support of more data source, sales prediction based on machine learning will promote the efficiency and value for enterprises. This paper aims at studying the significance of sales prediction based on machine learning, as well as analyzing the principle, advantages and limitations of three common machine learning approaches: support vector machine, XGBoost and random forest.