Detection of Financial Fraud by using Categorical Boosting based Bayesian Optimization

From the past few years, frequency of fraud attacks is targeting banking systems, financial institutions and credit card holders which highlighted the urgent need of advanced fraud detection systems to counter these threats. Traditional approaches for fraud detection had faced challenges which include feature over-load and high false positive rates. Therefore, this research proposes Categorical Boosting based Bayesian Optimization namely (CatBoost based BO) for detecting financial frauds. Initially, data is collected from dataset of credit card fraud detection which consists of European cardholders’ transactions made by credit cards in September 2013 respectively. Then, features are extracted by using Principal Component Analysis (PCA) to reduce dimensionality and eliminate redundancy. After that, CatBoost is utilized for detecting frauds and hyperparameter tuning is done by using BO which explores and selects optimal hyperparameters by learning from past evaluations, enhancing model performance. The proposed CatBoost based BO effectively achieved better results in terms of accuracy (0.9997), recall (0.9618), precision (0.7970) when compared with existing CatBoost respectively.