ATM Maintenance

The current scenario in the banking sector is showing a rapid decline in the number of bank branches with simultaneous increase in the number of ATM bases setup by each bank. However, inefficient maintenance of ATM bases by respective banks lead to longer downtimes thus causing customers to prefer a new bank. To reduce the loss of customers, banks should opt for predictive maintenance of their ATMs. In today’s world it is no longer enough to respond to outages when and after they occur, since this increases the down time of the ATMs and prevent customers from receiving services when needed. Predictive maintenance will enable proactively deciding when a maintenance visit is needed thus preventing unexpected ATM failures.

We planned to combine data on ATM logs, service history and environmental factors to form a model for predicting ATM outages. We extracted features from the log data to understand which thresholds and frequencies were reached for an outage to occur. The service data helped us identify certain error combinations that resulted in outages or failure of the ATMs. After the feature selection, we ran an analytical model to predict outages and determine a pre-defined time interval right before the occurrence of a failure.

The project implementation and documentation can be found here .