Analysis of machine learning techniques for credit card fraud detection

Rapid and easy transaction through credit card system has increased fallacious cases everywhere. Machine Learning algorithms has been applied for identifying fraudulent transactions. Fraud detection problem have two major issues, firstly, the legitimate and fraudsters consistently change behaviour and secondly, datasets are severely skewed. Implementation of the system littered with machine learning algorithms, excerpting variables and sampling accession on datasets. This paper investigates the performance of Logistic Regression, Random Forest, Decision Tree and SVM (Support Vector Machine). The proposed system shows results according to accuracy, sensitivity, specificity, precision of above techniques. Transactions in the dataset are heavily right skewed. Undersampling and oversampling is conducted on the data. The work is implemented in Python.