An Original Approach to Identify the Better Accuracy in Credit Card Fraud Transaction by Comparing Logistic Regression with K-Nearest Neighbours Algorithm

This work was extracted from the raw data by suggesting machine learning techniques like the clustering model K-Nearest Neural Network (KNN) to determine whether the input data classifies legitimate or fraudulent transactions and to obtain greater accuracy. Materials and Methods: Detect the fraud techniques utilizing the F1-Measure score is the mean between precision and recall. The range of F1 scores is 0,1. Implementing Machine learning algorithms, the sample size for Logistic Regression (N = 20), proposing the under-sampling technique is K-NN (N = 20) and G power (value = 0.8). Results: K-Nearest Neural Network method classifies any new incoming transaction calculating the K-Nearest point to achieve this accuracy. The independent sample T-Test (=.272) result (p0.05) with a 95% confidence level does not statistically support the two algorithms KNN and LR. Conclusion: K Nearest Neural Algorithm proves (99.4%) to detect fraud, false alert rate with accuracy it appears to be better than Logistic Regression (99.1%).