1Dr.M. RAMU, 2SIRI SATWIKA POLOJU,3 PUPPALASRINIJA, 4P KRANTHIKUMARREDDY 03 05 The rapid growth of digital payment systems ecommerceplatforms and online banking hassignificantly increased the volume of financialtransactions thereby elevating the risk offraudulent activities Traditional fraud detectionsystems primarily based on static rulebasedapproaches are no longer sufficient to handleevolving and sophisticated fraud patterns Thisproject presents a CloudBased Fraud DetectionSystem designed to accurately identify fraudulenttransactions using machine learning techniquesThe system incorporates data preprocessing featureengineering and classification algorithms toanalyze transaction patterns and distinguishbetween genuine and fraudulent activities Varioussupervised learning models including LogisticRegression Decision Tree Random ForestSupport Vector Machine Naive Bayes and KNearestNeighbors are implemented and evaluatedusing performance metrics such as accuracyprecision recall F1score and ROCAUC Toaddress the issue of imbalanced datasetstechniques such as resampling and SyntheticMinority Oversampling Technique SMOTE areapplied The selected optimized model is integratedinto a Flaskbased web application that enablesrealtime fraud prediction Cloud deploymentensures scalability flexibility and high availabilityallowing the system to handle large volumes oftransaction data efficiently The proposed systemenhances detection accuracy reduces falsepositives and provides a secure and reliablesolution for online transaction monitoring Overallthe integration of machine learning with cloudcomputing offers a robust framework forpreventing financial fraud and improving trust indigital financial systems Copyright (c) Journal of Science Engineering Technology and Management Science. All rights reserved 2026