Online Payment Fraud Detection Using Decission Tree and LSTM Neural Network

In today’s digitally connected world, online payments have become the backbone of financial transactions, leading to a parallel surge in fraudulent activities. The project titled “Online Payment Fraud Detection Using Machine Learning” focuses on designing and implementing a system that leverages advanced ML techniques to detect fraudulent transactions in real-time. Using Python-based tools and widely accepted datasets [14], [15], [17], this system incorporates supervised models such as Random Forest and Neural Networks to identify and classify fraudulent behaviours based on transaction data. To supplement core datasets, external insights were collected using the Google Trends API [2], [5] to incorporate macro-level behavioural indicators and user sentiment trends associated with fraud. The dataset is pre-processed using Pandas and NumPy for null handling, normalization, and outlier treatment. Synthetic Minority Oversampling Technique (SMOTE) is used to balance the class distribution [11]. Matplotlib and Seaborn are employed for visual exploration and feature selection. The ML models are optimized using cross-validation and hyperparameter tuning techniques, and the results are evaluated through metrics like Precision, Recall, F1-Score, and ROC-AUC [12]. The system is designed to scale, with future integration into platforms like Streamlet or Power BI for deployment in production environments. Real-world applications include fraud detection in banking, e-commerce platforms, mobile wallets, and insurance. By merging behavioural data, transactional attributes, and scalable ML architectures, this project builds a proactive and interpretable framework for combating online payment fraud