Machine Learning Model Selection for Banking & Insurance Applications Using Performance-Based Evaluation Metrics
Abstract: Selecting an appropriate machine learning model for banking and insurance applications is a critical research challenge due to regulatory constraints, data imbalance, and performance variability across datasets. Although several algorithms such as Logistic Regression, Support Vector Machines, Decision Tree, Random Forests, and Gradient Boosting Classifier are frequently applied in financial analytics, no universal method exists for selecting an optimal model. This paper presents a performance-driven model selection framework that uses quantitative evaluation metrics including F1-score, Area Under the Curve (AUC), precision, recall, and accuracy to empirically determine the most suitable model for financial classification tasks. Experiments are conducted on a real-world financial dataset to evaluate model reliability under imbalanced data conditions. The study demonstrates that model choice must be dataset-specific and guided by objective performance metrics rather than theoretical assumptions.
Keywords: Machine Learning, Model Selection, Banking Analytics, Insurance Risk Prediction, F1- score, AUC, Classification Metrics
How to Cite?: K Anantha Lakshmi, S Prashanth, M Narendra, "Machine Learning Model Selection for Banking & Insurance Applications Using Performance-Based Evaluation Metrics", Volume 14 Issue 12, December 2025, International Journal of Science and Research (IJSR), Pages: 1803-1807, https://www.ijsr.net/getabstract.php?paperid=SR251222132524, DOI: https://dx.doi.org/10.21275/SR251222132524