CREDIT CARD FRAUD DETECTION

Financial fraud is a major worry in the current digital era since technological improvements have made transactions easier while creating new criminal activity opportunities. To reduce losses, preserve confidence, and protect the general stability of financial institutions, fraud detection in real-time is essential. We will be able to examine enormous volumes of transaction data to find trends and anomalies connected to fraudulent conduct by utilizing the power of machine learning algorithms such as Logistic Regression, SVM, SGD classifier, KNN classifier, Naïve Bayes, Decision tree, Extra tree classifier, random forest, Extra trees classifier, MLP classifier and stack based model(DT, RF, Extra trees classifier).

CLASS IMBALANCE HANDLING

Address the issue of class imbalance in the dataset, as fraudulent transactions are usually a small proportion of the total data. Techniques like oversampling, undersampling, or generating synthetic data (e.g., SMOTE) can be used. When machine learning is used to detect credit card fraud, handling class imbalances is important to ensure the model’s effectiveness in detecting fraudulent transactions and the best approach depends on a particular dataset and machine learning algorithm chosen. Under-sampling_ seeks to balance the class distribution by randomly removing samples from the majority group (a non-deceptive task)._ Over-sampling is the replication of samples from a subclass (the use of fraud) to balance the class distribution. SMOTE (Synthetic Minority Over-Sampling Method) creates artificial samples for minorities by interpolating between existing minority class samples. In practice, there is no one-size-fits-all solution, and the best approach may vary depending on the specific dataset and machine learning. for the given dataset, it is preferred to use SMOTETomek for handling Class imbalance along with the Extra trees classifier model. Although handling class imbalance is important, If you neglect class imbalance it is preferable to use a stack model consisting of DT, RF, Extra trees classifier