Machine Learning for Anomaly Detection
- Posted by Wilker Amorim
- On December 8, 2019
- AI, Communications, Machine Learning, Symphony Platform, Telecommunications, Time Series Forecast
In this case, the objective was to build and implement an AI solution to monitor, analyze and alert Communications companies of:
- Deviations from normal traffic behavior
- Detecting robocalling
- Fraud.
The Machine Learning model used a Time Series structure applied to a large data pipeline based on aggregated data volumes from different sources. Continuous monitoring and analysis for anomalies generated results that automatically alerted the system. This enables new KPI measures to be developed that report automatically inside the product dashboards, at near real-time (5-15min) intervals. Daitan’s data science team worked with product/market domain experts and engineers in 3 different geographies to bring the project from inception to production release.