Improved Accuracy in Detection of Fraud Websites using Convolutional Neural Network Algorithm with Random Forest Algorithm
M. Karthick Roshan, S Loganayagi · 2024
This study aims to enhance fraud website detection through the utilization of the Novel Convolutional Neural Network Algorithm and the Random Forest Algorithm. The Novel Convolutional Neural Network Algorithm and Random Forest Algorithm are employed with diverse training and testing splits to predict the identification of fraud websites. While the Random Forest Algorithm demonstrates superior accuracy in predicting extremist reviewer groups in e-commerce, the Novel Convolutional Neural Network Algorithm, alongside a dataset of $\mathrm{N}=11053$ samples, offers a nuanced perspective by categorizing extremist reviews into distinct grounded categories. The Gpower test utilized achieves an approximate $85 \%$ accuracy with the following setup parameters: $\boldsymbol{\alpha}=0.05$ and power $=0$.85. The Novel Convolutional Neural Network Algorithm ($\mathbf{(6 6 . 1 0 \%}$) outperforms the Random Forest Algorithm ($\mathbf{9 5 . 2 3 \%}$) in accuracy. An independent samples t-test indicates a significant difference between the two algorithms, with a significance value of 0.04 (p $\lt0.05$). Conclusion: The accuracy of the Novel Convolutional Neural Network Algorithm surpasses that of the Random Forest Algorithm in detecting fraudulent websites. Novel Convolutional Neural Network Algorithm, Deception, Fraud Detection, Random Forest Algorithm, Swindling, Social Protection.