Improving Decision Tree Accuracy through AdaBoost Ensemble with SMOTE Oversampling and ExtraTreeClassifier Feature Selection

The issue of detecting credit card fraud is prevalent in today's society due to the increasing reliance on the internet and the rise of online transactions and e-commerce platforms. Fraudulent activities occur when cards are stolen or when unauthorized individuals exploit credit card information. To address this problem, a credit card fraud detection system with high accuracy is developed using the Decision Tree algorithm. The main focus of this study is to address data imbalance and identify the most influential features using the ExtraTreeClassifier algorithm. Additionally, a combination of Decision Tree and AdaBoost is employed to enhance the accuracy of fraud detection. The effectiveness of these algorithms is measured through the use of metrics such as accuracy, precision, recall, and F1-score. The algorithms' effectiveness is further assessed through the examination of ROC AUC curves. Through the process of comparing the performance of the Decision Tree and the combined approach with AdaBoost, we can determine the most effective method for detecting fraudulent activities based on the algorithm that demonstrates the highest accuracy, precision, recall, and F1-score. This comparison allows us to identify the optimal model for effectively identifying and preventing fraudulent activities in various domains, providing valuable insights to mitigate potential threats.