H Yperparameter Tunning to Improve Deep Learning Model Performance

Deep learning models have become a powerful tool in various domains, including fraud detection in credit card transactions. In this paper, we explore the crucial aspect of hyperparameter tuning to enhance the performance of deep learning models in the context of credit card fraud detection. We investigate four different combinations of optimization algorithms (Adam and Adagrad) and weight initializers (Xavier and He) with and without batch normalization. Through comprehensive experimentation using Convolutional Neural Networks (CNNs) on a credit card fraud detection dataset, we find that the combi-nation of Adam optimization with batch normalization and the He initializer consistently yields the highest accuracy compared to the other configurations. Our study not only highlights the significance of selecting appropriate hyperparameters but also provides insights into achieving superior performance in fraud detection tasks.