Machine Learning-based Oddity Detection of Smoke and Gas Sensor Data in a Large Gated Community

Oddity detection identifies strange events or observations that are statistically distinct from the rest of the data and can cause concern. The paper proposes a method to find oddities from an extensive dataset made up of data from smoke and gas sensors of the large gated community to identify a potential hazard (such as a fire) in the home. The dataset was created using a Python Script, and oddities were found using machine learning techniques with the PYCARET module. By applying an oddity score to each data sample, methods for detecting anomalies based on monsters, namely, the Local Outlier Factor (LOF) and the Isolation Forest (I-Forest) algorithms, are used. The entire program was developed using Google Colaboratory (also known as Colab), an open-source cloud service.