A Novel Accuracy Analysis of Credit Card Fraud Detection Using ResNet50 Over Linear Regression
Gurram Bharath, P Priyadarsini · 2024
A novel accuracy identification of fraudulent activity on credit cards using ResNet50 over linear regression (LR). For the purpose of detecting credit card fraud in online transactions, data mining was crucial. In our research, we will be using ResNet50 as proposed, which will be compared with existing algorithms such as linear regression (LR). The data set needed for the detection of credit card fraud is accomplished by utilizing ResNet50 and Linear Regression (LR) algorithms. For two algorithms, there are two groups each with 15 dataset sample size values and G power values as 75% and 95% confidence intervals. The accuracy of ResNet50 is $\mathbf{9 5. 3 9 \%}$ and the loss is 4.61%, which outperforms linear regression (LR) at $\mathbf{9 0. 7 6 \%}$ and loss at 9.24%, respectively. a T-test with independent samples and obtained a p-value of 0.00 ($\mathbf{p}\lt0.05$), where the level of significance is lower than 0.05. This implies the results are statistically more significant. The result shows that ResNet50 is statistically more significant than the linear regression for novel accuracy analysis of credit card fraud regarding accuracy