Anomaly detection model
Aitic partnered with a large Fintech firm to launch an anomaly detection model – NEWS (Network early warning system) that monitors the reason for customers calling to the customer care/support team to service their accounts and alerts to the teams when there is an unusual volume. This helps the organization to proactively identify events or news, including but not limited to fraudulent activities and unplanned events, impacting their customers and offer better service. The alert is also monitored by the Operation and Emergency response team to get alerted on the issues for timely handling and response.
To deliver the solution, our team of Data engineers, analysts, Data scientist and Product managers analyzed the data flow and partnered with several teams to create a pipeline for data collection and processing, and train and evaluate the best models – Statistical Process control, Moving average and Bayesian. The model output was rendered in the user interface in the time series format that has configurable parameters for individual user needs. Aitic also partnered with 3P news/events companies to map the related news to the anomaly identified.
Outcome
NEWS became the first place of notification for unusual events for the organization, now widely used by more than 10,000 users in the organization. The model has identified multiple events/news impacting their customers including data leak event, event cancellation, cyber attack news and political demonstration – to name a few. This feature has improved the customer and employee sentiments.