Tree Based Models: A Comparative And Explainable Study For Credit Default Classification

Abstract

Most of the real-world problems can be solved with the help of classification. Some problems that might feel to be regression can also be solved using classification by binning the values. But due to the presence of a large number of classification algorithms, it becomes difficult to choose one particular model. SVC, KNN, and Naive Bayes are some of the traditional models available which do the task quite efficiently, but when it comes to the explainability of the outcome these models fail to address them. This is when the real use of tree-based models comes in handy as we can visualize the entire flow of conditions that leads to a particular outcome. In this paper, we have tried to compare the performance and explainability of the Decision tree, Random Forest, AdaBoost, GBoost, and XGBoost for this we have considered the credit default dataset from Kaggle.