COMPARATIVE ANALYSIS OF PSO AND FIREFLY OPTIMIZATION FOR VIOLENCE REPORT CLASSIFICATION

Cases of violence against children and women continue to increase, but the handling of reports is often hampered by the large volume of incoming reports and the lengthy manual classification process. This study aims to address these issues by developing a method for automatically classifying reports of violence using the Support Vector Machine (SVM) algorithm optimized with Particle Swarm Optimization (PSO) and Firefly algorithms. The main objective is to group types of violence accurately to facilitate faster and more effective identification and handling. The research dataset consists of 500 reports obtained from Kaggle, with stages including text pre-processing, implementation of optimization algorithms, and evaluation based on accuracy, precision, recall, and misclassification error. The experiments were conducted using Python on the Google Colab platform. The results showed that PSO-SVM achieved an accuracy of 87.00% and a recall of 80.42%, outperforming Firefly-SVM which achieved an accuracy of 86.00% and a recall of 78.75%. Although Firefly-SVM demonstrated slightly higher precision (92.63%) compared to PSO-SVM (91.53%), PSO-SVM had a lower misclassification error (13.00% compared to 14.00%). These findings indicate that PSO-SVM is more effective for applications requiring better case detection, while Firefly-SVM is more suitable for applications prioritizing precision in positive predictions.