Online payments fraud prediction using optimized genetic algorithm based feature extraction and modified loss with XG boost algorithm for classification

Online payment fraud become a pressing concern in the digital age, necessitating robust predictive models to identify fraudulent transactions effectively. This research proposes a novel approach that leverages an Optimized Genetic Algorithm (GA) for feature extraction and a Modified Loss function in conjunction with the XGBoost algorithm for classification. The first step involves the application of a GA to optimize feature selection. Genetic algorithms mimic the process of natural selection, iteratively evolving a population of potential feature subsets to maximize the predictive power of the model. This optimization process helps identify the most relevant features for fraud detection, reducing dimensionality and enhancing model efficiency. Next, a Modified Loss function is introduced to the XGBoost algorithm. Traditional loss functions aim to minimize prediction errors, but they may not be directly suited for fraud detection, where the focus is on correctly classifying fraudulent transactions. The Modified Loss function is tailored to prioritize the identification of fraudulent cases, thus improving the model's ability to differentiate between legitimate and fraudulent limitations transactions. The proposed approach is evaluated using real-world online payment transaction datasets, and its performance is compared to traditional methods. Experimental results demonstrate the superiority of the optimized genetic algorithm-based feature extraction and the Modified Loss with XGBoost algorithm for classification in terms of fraud detection accuracy, precision, and recall. By improving the accuracy and efficiency of fraud detection systems, this methodology can help financial institutions and e-commerce platforms protect their customers from fraudulent activities while reducing false positives.