A Comparative Analysis of the Performance of Decision Trees, Random Forests, and XGBoost in Classifying SAR Information Production Priorities

Abstract

Advances in information technology have driven the need for fast, accurate, and targeted management of SAR information. In the information production process, prioritization is crucial so that urgent information with widespread impact can be produced first. This study aims to analyze and compare the performance of the Decision Tree, Random Forest, and XGBoost algorithms in classifying the priority of SAR information production. The data used consists of 500 synthetic data points organized based on SAR information production scenarios with three priority classes: low, medium, and high. The attributes used include information type, content type, publication platform, source material, operational status, contextual location, urgency level, public impact, deadline, completeness of material, production complexity, need for revision, estimated production duration, and number of available personnel. The research stages included collecting synthetic data, preprocessing the data, splitting the data at an 80:20 ratio, training the model, and evaluating the model. Evaluation was conducted using accuracy, precision, recall, F1-score, and a confusion matrix. The results showed that Random Forest achieved the best performance with an accuracy of 98.00%, precision of 98.03%, recall of 98.00%, and an F1-score of 98.00%. XGBoost achieved an accuracy of 97.00%, while Decision Tree achieved an accuracy of 93.00%. These results indicate that Random Forest is more effective at classifying SAR information production priorities using the synthetic data provided. The findings of this study can serve as a preliminary framework for utilizing machine learning to support a more objective classification of SAR information production priorities.