HasLoss: A Novel Hassanat Distance-Based Loss Functions for Binary Classification

Ahmad S. Tarawneh^*

• Mutah University, Al Karak, Jordan

Loss functions play a critical role in machine learning, particularly in training neural networks for classification tasks. In this work, we introduce a novel adaptation of the Hassanat distance and formulate six variants employed as loss functions, designed specifically for binary classification. Their effectiveness is evaluated on synthetic datasets and nine real-world datasets, ranging from a few hundred to nearly 48,000 samples, under controlled experimental conditions. A comprehensive comparison is conducted against widely used loss functions, including Binary Cross-Entropy (BCE), Focal Loss, Mean Squared Error (MSE), and L1 Loss. Results show that the proposed Hassanat-based losses achieve competitive performance across evaluation metrics, and on some datasets they surpass existing methods in calibration, convergence speed, precision, recall, F1-score, and AUC, while also exhibiting robustness to outliers and noise. Importantly, Cohen's d effect size analysis shows that some of the proposed variants introduce a larger practical effect size than popular loss functions such as BCE. This study highlights the potential of distance-based loss functions and establishes a foundation for further exploration in robust model training.