Financial fraud detection represents a critical challenge for financial institutions due to the complex, evolving, and costly nature of fraudulent transactions. Although AdaBoost has demonstrated strong performance on imbalanced fraud datasets, its effectiveness is constrained by hyperparameter sensitivity and susceptibility to overfitting. This study presents a financial fraud detection system based on an AdaBoost ensemble model optimized by the Starfish Optimization Algorithm (SFO). The SFO, a recently developed bio-inspired metaheuristic that simulates the exploration, preying, and regeneration behaviors of starfish, is applied to automatically tune the key AdaBoost hyperparameters: number of estimators, learning rate, and maximum tree depth. The study follows a structured machine learning pipeline using Kaggle’s Credit Card Fraud Detection dataset comprising 10,000 anonymized transaction records. Preprocessing involved missing value imputation, outlier treatment using Z-score and Interquartile Range (IQR) methods, feature selection and engineering, categorical encoding, and Min-Max normalization. A baseline AdaBoost ensemble model was first constructed using decision stumps as weak learners, followed by SFO-based hyperparameter optimization through a defined fitness function. The optimal parameters were then used to retrain the classifier, producing the SFO-AdaBoost model. The model was implemented in MATLAB R2023a and evaluated using a 70/30 training–testing split with random subsampling cross-validation. Experimental results demonstrated that the SFO-AdaBoost model achieved a false positive rate of 6.67%, sensitivity of 97.00%, specificity of 93.33%, precision of 97.14%, F1-score of 96.52%, accuracy of 95.90%, and a detection time of 24.50 seconds. These results represent clear improvements over the standard AdaBoost model across all evaluation metrics. The findings confirm that the Starfish Optimization Algorithm provides an effective and efficient mechanism for automating AdaBoost hyperparameter tuning, resulting in a robust, scalable, and computationally efficient fraud detection framework.

Keywords: Financial Fraud Detection, AdaBoost, Starfish Optimization Algorithm (SFO).