Empirical Study: Machine Learning Analytics of Automated Teller Machines Data for Enhanced Customer Satisfaction and Banking Service Integrity in Nigeria

Automated Teller Machines (ATMs) are ubiquitous in modern banking and generate large volumes of data on user behavior and preferences. This study analyzed big data from ATMs to identify patterns in ATM usage, observe potential fraud, and optimize ATM placement and maintenance. Data were collected from several banks in Nigeria; covering over 2 million ATM transactions conducted over a one-year period, including device status logs. Our analysis reveals several key findings, including the importance of proximity to users' home, work, and market locations in ATM placement, the devices responsible for service downtime, the possibility of fraudulent transactions involving identity theft, and the potential for personalized marketing campaigns based on user demographics and transaction history. ATM patronage during active hours was observed between 9 am and 7 pm, with cash withdrawal being the major transaction for most customers, typically ranging between ₦2,000 and ₦40,000. Additionally, network fluctuations (65.8%), card reader issues (18.2%), cash jams (11%) and PIN pad (5%) issues were identified as major contributors to ATM downtime. Lastly, ATM fraud typically occurs outside customers’ proximity and transaction patterns. These insights have important implications for banks and other financial institutions seeking to improve their services and maximize customer satisfaction. Keywords: ATM, Machine Learning, Unsupervised Learning, Big data, Transaction Monitoring Ojulari, H.O., Oke, A.O. & Arulogun, O.T. (2024): Empirical Study: Machine Learning Analytics of Automated Teller Machines Data for Enhanced Customer Satisfaction and Banking Service Integrity in Nigeria. Journal of Advances in Mathematical & Computational Science. Vol. 12, No. 1. Pp 87-102. Available online at isteams.net. dx.doi.org