URL prediction, Random Forest Malicious URL Prediction Using Classification Approach with Random Forest and LightGBM Algorithms

Abstract

This study aims to develop a method for predicting malicious URLs using Random Forest and LightGBM algorithms. The dataset used in this research comprises 651.191 URLs categorized as benign, defacement, phishing, and malware. Key features such as URL length, number of directories, and URL abnormality are extracted and used to train the models. Evaluation results indicate that both algorithms have high accuracy in classifying URLs based on the provided features. The program allows users to input URLs for evaluation and provides predictions on the URL categories. The developed method is expected to enhance digital security by providing an effective prediction tool against malicious URL threats.