An Efficient Approach to Detect Fraudulent Service Enrollment Websites with Novel Random Forest and Compare the Accuracy with XGBoost Machine Algorithm

Abstract

Aim: The main aim of this research study is to detect fraudulent service enrollment websites using the Novel Random Forest algorithm and compare its accuracy with the XGBoost classifier algorithm. Materials and Methods: This research involved comparing two groups namely Random Forest and XGBoost. In this study, 1784 dataset samples had been utilized for statistical analysis. Dataset splits into training and testing which have 1200 of training and 584 of testing. The Gpower test was utilized with a setting parameter of 85% (α=0.05 and power=0.85) to determine the appropriate sample size. With a sample size of 10 and a confidence interval of 95%, we aimed to predict fraudulent service enrollment websites. Results: The significant value of p=0.000 (p<0.05) is statistically significant for detecting fraud websites. The Novel Random Forest algorithm demonstrates higher accuracy in recognizing objects and enhancing the evaluated data, with an accuracy rate of 92.634%, compared to the XGBoost classifier which achieves an accuracy of 75.545%. Conclusion: The accuracy of Novel Random Forest is better when compared to accuracy of XGBoost classifier.

Keywords

• novel random forest
• xgboost classifier
• machine learning algorithms
• fraud detection
• website detection system
• neural network
• social protection