Machine Learning in IoT for Smart Home Automation

Abstract

Automation and AI are pivotal across various domains, with the primary goal of the proposed AI-Based Smart Geyser being power conservation and device automation aligned with user usage patterns. The machine learning model is constructed using the K-Nearest Neighbors (KNN) algorithm and SciPy Toolkit (SCIKIT) and is integrated with an IoT backend deployed on the cloud. Within this application, the system forecasts whether the device should be activated or deactivated based on the user's historical usage metrics at specific times. Utilizing the Node MCU, the device connects to the cloud via WIFI, while the sequence of operations is managed by the Arduino UNO according to user preferences. The proposed system demonstrates an accuracy exceeding 95%, which may vary depending on the volume of data used to train the AI model. Additionally, evaluating the AI recommendation system's performance in predicting device states and determining the timing using a Geyser model reveals precise predictions aligned with user usage patterns.