An Empirical Study on Bankruptcy Prediction using Ensemble Learning

Abstract

Bankruptcy prediction helps to assess the financial condition of a company and its future perspectives within the context of long-term operation on the market. Using machine learning to solve this problem can be a time-efficient and cost-effective approach. This paper introduces an approach using Ensemble learning methods to tackle the bankruptcy classification problem, which achieved the fourth position on the leader board of The 3rd Annual International Data Science & AI Competition 2022 - Structured Data Track. The data set given by the committee of the competition has a lot of challenges so we perform some preprocessing and feature engineering techniques to make the data set become cleaner for modelling. We use three ensemble algorithms, namely Random Forest, Catboost, and LightGBM to compare the performance of three algorithms on the bankruptcy classification problem and find the best result to submit to the competition. After experimenting, we achieve the best result at 98.21% Accuracy on the private leader board of the competition. The result comes from the LightGBM model trained on the data set which is enhanced through feature engineering techniques.