Immune-inspired Anomaly Detection In Environmental Sensor Data with Natural Language Generation Support

IADES is a partnership project funded by RCUK dot.rural Digital Economy Hub, Weather2 ltd, and Data2Text ltd. Its aim is to develop a prototype framework inspired from human immune system to detect anomalous patterns from high resolution environmental sensor data of multiple sources.

Principle Investigator: Dr. Wei Pang (Aberdeen)

Co-investigator: Prof. George M. Coghill (Aberdeen)

Co-investigator: Kit Macleod (James Hutton)

Feasibility study of developing an immune-inspired anomaly detection framework for environment sensor data.

Make the anomaly detection framework ready for NLG techniques, which could achieve better communication of environmental anomalies.

Investigate how to extract spatial-temporal patterns using immuneinspired approaches from environment sensor data.

Develop a prototype model for identifying anomalies from historical data.

In this research we aim to build a software tool to detect abnormal behaviour in the environment from measured data. This can help us predict natural disasters (e.g. floods) and extreme weather events. The design of this tool will be inspired by the human immune systems: similar to the way that the immune systems protect us from foreign invaders, this tool can detect potential threats to the environment. To make users better understand the abnormal behaviour in the environment, we will make the tool ready for future integration with natural language generation (NLG) techniques, which can generate natural language from the output of the detection tool.

In this project we use river levels data from SEPA (Scottish Environment Protection Agency) and hydrological and meteorological data from our industrial partners (Data2Text Ltd and Weather2 Ltd) to test our detection tool.