Detection of Tax Arrears Based on Ensemble Leaering Model

Abstract

Machine learning technique has been widely applied in many applications, e.g., stock prediction and image classification. In this paper, we construct an ensemble model to detect whether there are tax arrears in enterprises. Tax department can use this model to detect tax arrears in advance, avoiding tax arrears. The ensemble learning model consists of six base classifiers, i.e., Multi-Layer Perceptron(MLP), k-Nearest Neighbor (KNN), Random Forest(RF), Extremely randomized Trees (ET), Gradient Tree Boosting (GTB) and XGBoost. Soft voting with weight is used to combine the base classifiers. Experimental results show satisfying performance of the proposed method on the tax dataset of N anhai, Foshan, China in 2015 and 2016.