Credit Card Fraud Detection comparing multiple Supervised Learning Algorithms for Optimal Accuracy

Link to dataset: https://www.kaggle.com/datasets/nelgiriyewithana/credit-card-fraud-detection-dataset-2023

Order of operations:

1. MP Stage 1:

• 12 individual Supervised Learning algorithms are trained, tested and compared across 4 factors on the same dataset.
• The algorithms are ExtraTrees Classifier, Perceptron, AdaBoost Classifier, Naive Bayes, Stochastic Gradient Descent, Extreme Gradient Boost, Multi Layer Perceptron, Random Forest Classifier, DecisionTrees Classifier, K Nearest Neighbour, Support Vector Machine and Logistic Regression Model.
• The comparison factors are Fitting Time, Training Accuracy, Testing Accuracy and Overall Accuracy.
• The Models which passed Stage 1 and moved on to Stage 2 are Stochastic Gradient Descent, Perceptron, AdaBoost Classifier, Naive Bayes and K Nearest Neighbour.

2. MP Stage 2:

• 5 of our original 12 Supervised Learning Algorithms are trained, tested and compared across a greater set of Performance Evaluation Metrics.
• The Performance Evaluation Metrics are Training Accuracy, Testing Accuracy, Optimal Testing Accuracy, Sensitivity, Specificity, Precision, F1-Score, Matthews Correlation Coefficient, Area Under the Receiving Operating Characteristics Curve, Balanced Classification Rate, Average Precision, Choen Kappa, Log Loss, Training Time and Optimal Threshold.

3. MP Testing:

• Base Level Unit Testing, Functional Testing and Integration Testing is done for the all the models to check if they will work the same in the specific environments for their deployment or implementation and ensure Robustness and Reliability.

4. MP Comparison Result:

• The various levels of testing and comparison gives us 2 good Models at the top, which alows us to take a Hybrid approach to create a Model with the best qualities of both Stochastic Gradient Descent and AdaBoost Classifier. This Hybrid model has been labelled as Hybrid1.
• The Hybrid1 model is compared against SGD and AdaBoost and the results are shown in graphs.

Research Available at: https://www.researchgate.net/publication/390299264_Credit_Card_Fraud_Detection_comparing_multiple_Supervised_Learning_Algorithms_for_Optimal_Accuracy