Statistical Assessment of Bayesian Optimization Gradient and Hessian Computation Based Improved Random Forest Classifier for Non-Linear Classification Problems

Abstract

This study introduces an enhanced Random Forest classifier, optimized using a novel Bayesian optimization approach that incorporates gradient and Hessian computations. Our objective was to improve the model's accuracy and computational efficiency when applied to this specific type of image data. We conducted a series of experiments to statistically assess the performance of our proposed classifier against conventional models. Using a robust dataset of annotated images depicting various stages of lumpy skin disease, our model demonstrated superior performance in terms of accuracy, sensitivity, and specificity. Bayesian optimization effectively tuned the hyperparameters of the classifier, leading to significant improvements in learning rates and decision boundary formations. This paper details our methodology, experimental setup, and statistical validations, highlighting the benefits of our approach. Our findings suggest that the improved Random Forest classifier can serve as a powerful tool for veterinary diagnostics and may be adaptable for other complex image classification tasks.