Imbalanced data is a major challenge in anomaly detection domain, with huge non-anomalous data and limited anomalous data. This solution is capable of handling data imbalance and is a semi-supervised approach which uses generative deep learning algorithms. It learns normal IoT sensor patterns using non-anomalous data and builds a 1-rule threshold model using data from both classes. It then identifies the anomalous behavior of the sensor using inclusion-exclusion principle. The solution is also re-trainable to capture information drift.

This solution is capable of handling huge amounts of class imbalance and capturing anomalous behavior. This makes the solution usable in multiple industries for predictive maintenance to raise early alarms indicating system’s abnormal behavior. Specific use cases could be production machine maintenance, IoT sensor data analysis for anomalous behavior identification, IT infrastructure maintenance, data drift detection.

InfraGraf is a patented Cognitive infrastructure automation platform that optimizes enterprise technology infrastructure investments. It diagnoses and predicts infrastructure failures. Need customized Machine Learning and Deep Learning solutions? Get in touch!