Hybrid Random Forest and Naïve Bayes Models Optimized With Grasshopper Optimization Algorithm for Disc Herniation Prediction

Abstract

Intervertebral disc herniation is a prevalent spinal disorder that can lead to severe discomfort, neurological impairment, and reduced quality of life. Effective treatment planning depends on a timely and accurate diagnosis. In this study, we propose an advanced machine learning framework to improve the accuracy of disc herniation prediction by using the Random Forest Classifier (RFC) and the Naïve Bayes Classifier (NBC), both of which are optimized using the Grasshopper Optimization Algorithm (GOA). There are 500 patient records in the dataset, which includes imaging-derived parameters and clinical features. Training (70%) and testing (30%) subsets of the preprocessed data were separated. To enhance classification performance and adjust hyperparameters, the GOA was employed. Accuracy, precision, recall, and F1-score were used to evaluate the model. According to empirical findings, the hybrid RFC-GOA model performed better than any other model, with 91.5% accuracy, 92.1% precision, 90.4% recall, and 91.2% F1-score. With an accuracy of 90.4% as opposed to 85.1%, the NBC-GOA model also outperformed the baseline NBC model. These results demonstrate the efficacy of metaheuristic optimization and the superiority of ensemble-based approaches in medical classification tasks. In order to support clinical decision-making and enhance patient outcomes, the suggested models provide a reliable and understandable method for the early prediction of disc herniation. This study demonstrates the potential for creating trustworthy diagnostic tools for spinal disorders by fusing bio-inspired optimization techniques with machine learning classifiers