Improving the accuracy of Anomaly Detection in Multimodal Sensors using 1D-CNN
Muhammad Imad, Ian Cleland, Patrick McAllister, Chris Nugent · 2024
Unusual sensor data within intelligent built-up environments can indicate a range of concerns, including sensor inaccuracies, susceptibility to security breaches, and alterations in activity and behavioural patterns. This study aims to assess the effectiveness of 1D-CNN in detecting and improving the accuracy of anomalies in multimodal sensor data. This method effectively captures temporal patterns in lengthy data sequences collected over extended periods of time. Through comprehensive experiments utilising a public dataset for smart homes, we have empirically verified, after balancing the dataset, the proposed technique's efficacy, and a high accuracy of 0.96 in predicting anomalies.