XGBoost Approach For Transaction Prediction With Data From Bank

The gradient boosting(XGBoost)algorithm which based on decision tree is popular for solving classification or regression problem on big data set.As we known,XGBoost algorithm based on decision tree model,and the decision tree model cannot effectively identify the frequency features.However,these features contain hidden information in the original data.If the frequency features can be exploited,then the accuracy of prediction can be improved.The novelty of this article is to propose an improved XGBoost algorithm by exploiting the frequency features,thereby increasing the performance of the XGBoost on such specific data set.At the same time,this article also uses the specially designed neural network to train and predict on such specific data set,which also achieves the same accuracy.At the same time,this article will also compare with some classic machine learning algorithms to show its efficiency and feasibility of this improved algorithm.