The main goal of the project is to develop a hybrid machine learning model that effectively detects fraudulent financial transactions using a combination of Naïve Bayes and Support Vector Machine (SVM) classifiers. Fraudulent activity in financial systems poses serious threats to customers and institutions, often going undetected in traditional rule-based systems. This project addresses the problem by integrating probabilistic reasoning (Naïve Bayes) with boundary-based classification (SVM) to achieve better accuracy and reduce false positives. The final product will be a working prototype that can classify incoming transactions as either legitimate or suspicious, thus providing a critical tool in the fight against digital financial fraud.
This project follows a structured twelve-week timeline. In the initial weeks, students will set up their development environment using Python (Anaconda or Google Colab) and gather labeled and unlabeled datasets representing financial transaction patterns. They will explore and import required libraries such as Pandas, Scikit-learn, NumPy, and Matplotlib.
The middle phase of the project includes building a hybrid classification model using both Naïve Bayes and SVM, training it on curated data, and testing its performance against new transactions. Accuracy assessment, misclassification handling, and performance optimization will be key focus areas. In the final stages, students will refine the model, conduct system testing, complete the documentation, and deliver a final team presentation. Although the system is built to enhance fraud detection capabilities, its effectiveness depends heavily on data quality and proper training to prevent harmful misclassifications.
