Machine Learning-Based Empirical Investigation for Credit Scoring in Vietnam’s Banking
In thons for credit scoring in Vietnam with machine learning models based on our submissions for the Kalapa Credit Score Challenge. We conduct experiments with modern machine learning methods based on ensemble learning models: LightGBM, CatBoost, and Random Forest. Our experimental results are better than single-model algorithms such as Support Vector Machine (SVM) or Logistic Regression. As a result, we achieve the F1-Score of 0.83 (Random Forest) with the sixth place on the leaderboard. Subsequently, we analyze the advantages and disadvantages of the used models, propose suitable measures to use for similar problems in the future, and evaluate the results to select the best model. To the best of our knowledge, this is the first work of the field in Vietnamese banking.
PDF Abstract