Abstract: Effective and efficient fraud detection techniques are critical in the current surge in online transactions to ensure sustainable communities and cities. A promising approach to real-time fraud detection is the utilization of online predictions, which leverage machine learning (ML) algorithms to assess the likelihood of fraudulent activities. This chapter investigates the application of online predictions within a financial institution, specifically for detecting fraudulent transactions. Moreover, it delves into the data preprocessing techniques to adequately prepare the data for analysis, the ML algorithms utilized for prediction purposes, and the evaluation metrics adopted to measure the prediction model’s performance. Furthermore, this chapter thoroughly examines the practical implications of using online predictions for fraud detection, including the potential benefits, challenges, and prospects. By providing comprehensive insights into the use of online predictions for fraud detection, this chapter effectively underscores the critical role played by ML techniques in combatting fraud within online transactions and avails future perspectives and trends in fraud detection. Finally, the future of fraud detection using online predictions is poised for significant advancements. These trends will enable organizations to stay ahead of evolving fraud techniques, protect their assets and customers, and maintain trust and acceptance in an increasingly digital and interconnected world.