AI-Driven Fraud Detection in Insurance Claims: A Deep Learning Framework for FinTech Risk Intelligence

Abstract

The growth rate of fraudulent insurance claims is a significant challenge to financial institutions, and it is causing huge losses to the economy and inefficiency in operations. The purpose of this research is to find a solution to this growing problem by creating an automated system that can reliably forecast the number of fraudulent property insurance claims. To combat insurance data misclassifications and improve fraud detection, an Artificial Neural Network (ANN) machine learning model is trained in this study. The model has been trained and evaluated using a publicly available Kaggle dataset, which has 38 customer profiles, claim history, and policy details as features. The use of extensive preprocessing operations such as cleaning of data, label coding, working with missing values and balancing with the help of SMOTE, and normalization with the help of Standard Scaler provide quality of data and resilience. The proposed ANN model is tested with the help of common performance measures, including accuracy (ACC), precision (PRE), recall (REC), F1-score (F1), confusion matrix, ROC curve, and AUC. As the experimental outcome shows, with an accuracy of 96.67, the experiment is more accurate than the current baseline models like XGBoost, Decision Trees and Bi LSTM networks. The results show that ANN is effective in identifying unusual patterns of claims and this can be applied in enhancing the fraud prevention strategies and decision-making in the finance and insurance industry.