Anomaly Detection In An IoT-Acquired Environmental Sensor Data

Abstract

A demand in monitoring a landfill site is increasing. A landfill site produces methane which is a poisonous gas and harmful for the existence of nature. The monitoring of the landfill sites are made possible using sensors, but the data quality issues arising from sensors persist. This research focusses on solving the problem of anomaly detection, in turn, solving the issues related to Data Quality and giving an indication for the presence of a subtle anomaly. An exhaustive search to solve the problem and to apply the existing techniques is the core idea. This also involved speaking to other researchers from the industry. Basic techniques ranging from Gaussian Mixture Model to an Autoencoder is implemented. Finally, Various problems in finding a perfect solution are reported, and an ensemble approach to solving the problem of anomaly detection in ecological monitoring is proposed.