A Major Project-1 On Online Payment Fraud Detection using Machine Le: ing in Python. SUBMITTED TO RAJIV GANDHI PROUDYOGIKI VISHWAVIDYALAYA (M.P.) In Partial Fulfillment of the award of Bachelor of Technology in Computer Science and Engineering YEAR 2021-25 Submitted By: SHALILENDRA NIRANJAN ROLL. NO. 0905CS211158 SHIVANSH SAXENA ROLL. NO. 0905CS211163 TANUSH SAHU ROLL. NO. 0905CS211187 VIPIN YADAV ROLL. NO. 0905CS211199 Under the Guidance of: Mr. Nitin Dixit Assistant Professor Department of Computer Science and Engineering “THINK BIG... THINK BEYOND” INSTITUTE OF TECHNOLOGY & MANAGEMENTGWALIOR (M.P.) - 475001CANDIDATE DECLARATIO We SHATLENDRA NIRANJAN, SHIVANSH SAXENA. TANUSH SAHU and VIPIN YADAV of B.Tech. (Computer Science) VIII semester Roll No 0905CS211158, 090: O90SCS211187 and 090SCS211199 hereby declare that the Project entitled “ONLINE PAYMENYT FRAUD DETECTION U MACHINE LEARNING IN PYTHON” which is being submitted to Department of computer science & Engineering in ITM, Gwalior is our authentiework carried out in our VII semester. We declare that our work hes not been submitted in part or in full to any other university or institution for the award of any degree or diploma. SHAILENDRA NIRANJAN cooscsni sae SHIVANSH SAXENA (0905CS211163) ‘TANUSH SAHU (0905CS211187) VIPIN YADAV (0905CS211199),INSTITUTE OF TECHNOLOGY & MANAGEMENT GWALIOR (M.P.) - 475001 vom Visit us at website waywitmgoiin “AlN indlia Council for Technical Fal : ‘ Jyatay sytyoiks Vishwa CERTIFICATE This is to certify that the project entitled “ONLINE PAYMENYT FRAUD DETECTION USING MACHINE LEARNING IN PYTHON” being submitted by Shailendra Niranjan (Enroll, No. 0905CS211158), Shivansh S: ena (Enroll, No. 090SCS211163), Tanush Sahu (Enroll. No. 09058211187) and Vipin Yadav (Enroll, No, 0905CS211199) in partial fulfillment of the requirement for the award of B. Tech. degree in Computer Science & Engineering to Rajiv Gandhi Proudyogiki Vishwavidyalaya, Bhopal (M.P.) is a record of bonafide work done by them, under my guidance. 4 so . oe HEAD OF DEPARTMENT: PROJECT GUL Dr. Rishi Soni Mr. Nitin me Department of CSE Assistant rrotiy ITM, Gwalior (M.P) Department of CSE ITM, Gwalior (M.P) EXTERNAL EXAMINER PageNUTE OF TECHNOLOGY & MANAGEMENT We AUOR MD he AUS at website wan iuinodin 1 Vishysavilyal ACKNOWL “(DGEMENT result of AL Moment of suecessti i i montent of suecesstitl completion of an endeavor is always 2 plicitly or explicitly there in and it is impossible without the help and people around us, The suecessti fon of thi i syple around us, The successtinl completion of this thesis report is: efforts put in by many people and this report would have been plete without giving due credit to them, This acknowledgement is indeed a small ide in recognition of their help. our gratimde to Dr. Meenakshi Mazumdar, Director, Institute of and Management, Gwalior for her valuable suggestions. We feel highly to express our sincere thanks and deepest sense of gratitude to our guide Mr. Nitin Dixit who has spared her precious time & guided us to present this dissertation. y acknowledge her for extending her valuable guidance, support for literature, & critical review of report & above all the moral support she has provided us at all stages of to thank our Head of the department Dr. Rishi Soni, for his support and encouragement. Our thanks are extended to all staff members and persons who directly or indirectly helped us in achieving our goals. We could not have completed this work without the guidance, confidence and encouragement of our parents. They Geserve more credit than We can give for instilling in us a good work ethics and a desire to always learn more. Our thanks are also extended to all our friends who have kept our spirits high throughout this work. SHAILENDRA NIRANJAN Enroll. No. 0905CS211158 SHIVANSH SAXENA Enroll. No. 0905CS211163, TANUSH SAHU Enroll. No. 0905CS211187 VIPIN YADAV Enroll. No. 0905CS211199 achapter 1. Introduction 1.1 Aims, Objectives, Vision Chapter 2. Project / Problem Selection Chapter 3. Project Monitoring Syste 3.1 Gantt Chart a 3.2. PERT Chart Chapter 4. System Study 4.1 Existing System along with limitations/deficiencies 4.2 Proposed system along with intended objectives 4.3 Feasibility study 4.3.1 Operational 4.3.2 Technical 4.3.3 Economic Chapter 5. System Analysis 5.1 Requirement Specification 5.2 System flowcharts 5.3. DFDs/HIPOs/ERDs 5.4 Sources of data Chapter 6. Scope of the ProjectChapter 1. Introduction 1.1 Aims. Objectives. ision ~ Aims : The aim of this projec > Aim f this project is to design and implement an intelligent fraud detection system that can predict and identify fraudulent transactions from a dataset - > Objectives : Build a classification model to detect fraudulent transactions using machine learning algorithms. Use techniques like SMOTE and Tomek links to address class imbalance issues. Visualize data and prediction outcomes effectively for better insights. . Integrate tools like SMOTETomek for balancing the data and improving model performance. } Vision : The vision is to deploy a robust fraud detection system capable of continuously learning and adapting to emerging fraudulent behaviors, thus ensuring better protection for financial transactions.oblem Selection -.qudulent transactions ee reutions and their anni ee threat to financial Josses and damage to trust. The pS ees transactions and financial services has aan . increase in online to jdentify and prevent fraudulent netiviti 3 in ee a fraudsters continuously evolve their ta ste " mak aa increasingly difficult for traditional detection ee Paine up: tial monetary A major challenge in fraud detection is ‘the class imbalance problem, where the number of legitimate transactions far outweighs the number of fraudulent ones, often resulting in piased machine learning models that fail to detect fraud effectively. Traditional fraud detection systems rely on rule- pased methods that are often unable to adapt to new fraudulent techniques and may also produce a high rate of false positives, further complicating the detection process. Inadequate detection methods, along with the volume of transaction data, pose a significant problem. A more dynamic, data-driven approach is required to address these challenges effectively. The project aims to develop an intelligent fraud detection system based on machine learning techniques that: > Accurately classify transactions as legitimate or fraudulent. + Address the issue of imbalanced datasets using data preprocessing techniques like SMOTE (Synthetic Minority Over-sampling Technique) and Tomek Links to balance the dataset., Utilize advanced classificati Random Forest, XGBoost on F ‘cti Logisti a improve prediction accuracy. gistic Regression algorithms such aS Optimize model perforr ; 2 ance through : ics . such as accuracy, preci igh evaluation metr 4 “Y> Precision, recall, and F1-score. , Provide financial institutions with a reliable tool for fraud prevention, reducing the risks associated with online transactions. The proposed solution is designed to be scalable, adaptable to yarious data sources, and capable of improving over time through continuous learning. Ultimately, it aims to assist financial institutions in detecting and mitigating fraud more efficiently, minimizing financial loss, and enhancing customer trust.Chapter 3, Project Monitoring System 4.1 Gantt Chart nena nesses wane 1 ba a CT | oun ia eos Li) te Ly = @ fadbgeerunen 3.2 PERT Chart a itn co SE ence Val Veh ‘ —— » wi - es yO SLThe existing > High rates of f are classified as le > Imbalanced da s prefer the majority cla > Lack of real-time fr: systems. which lead to biased models that 1d detection capabilities in some 4.2 Proposed System Along with Intended Objectives The proposed system will address the limitations by: » Using machine learning algorithms like Random Forest and XGBoost for classification. > Implementing data preprocessing techniques like SMOTE and Tomek links to handle class imbalance. » Visualizing the results through accuracy metrics, confusion matrices, and ROC curves. 4.3 Feasibility Study 4.3.1 Operational Feasibility The system is operationally feasible as it focuses on implementing an algorithmic solution that can easily be integrated into existing infrastructure. The system can be deployed on cloud platforms, enabling fast processing of financial data. 4.3.2 Technical Feasibility The technical feasibility ‘has been confirmed, as the tools required, such as Python, Scikit-learn, Pandas, and SMOTE/Tomek links, are well-established and widely used in the industry. 10Economic Feasibility The cost of implementing this 4 : system is low since it relics on open-source libraries and existing hardware infrastructure. The penefits, in terms of fraud prevention and reduced financial losses. far outweigh the initial implementation cost.Chapter 5, System Analysis §.1 Requirement Specification ctional Requirements: > The system should be able to load and preprocess the ston data ® It should classify transactions into ‘fraud’ or ‘non- fraud? categories, » It must handle class imbalance using SMOTE and Tomek links, » The system should generate performance metrics such accuracy, precision, re E + Non-Functional Requiren » Tigh availability for » Scalability to handle larg 5.2 System Flowcharts5,3 Sources of Data The primary data source will be obtained from open datasets lik dataset. Data sources include: + Internal data repositories of financial institutions. + Publicly available datasets for testing and model evaluation, transaction datasets, possibly the Kaggle credit card fraud 13Chapter . UPLOr @, Scope ot the Project In Scope: ) process Data Pleprocessing and handling Classification Model de i Evaluation of the Mode » Visualiz oF imbalanced dé Velopment using machine le: ring. Lusi KEY performance metrics. atio! Presutts HON OF results and system performance Out of Scop: » x als me traud detection in production environments (This w iL be part of future developments). > Integrating the System with live transactional systems. ed machine learning techniques beyond basic ssification, Conelusion The project aims to create an effective fraud detection system using machine learning algorithms, overcoming common challenges such as_— class. imbalance = and_~—s modell underperformance. By implementing techniques like SMOTE and Tomek links, the system ensures better detection of fraudulent transactions, thus reducing financial losses for institutions. Future improvements include real-time fraud detection and further optimization of the machine learning models.