A Recommendation System Based on Extreme Gradient Boosting Classifier

Abstract

Shopping online has become the mainstream way of shopping of the society in recent years. Consumer behavior records on shopping website contain a lot of important information that can be used as basis for commodity recommendations. But traditional collaborative filtering-based recommendation systems are sometimes difficult to handle noise in behavior records. In this paper, we proposed a complex model based on eXtreme Gradient Boosting(xgboost) algorithm with a series of methods of features extraction to build a recommender system based on behavior records of consumers got from Alibaba mobile client. The system had a good performance on the data set with f'1-score of 7.97% and has high time efficiency.