AHEACS: A Streaming Dataset for Automatic Home Electrical Appliance Control Systems

Abstract

Introducing AHEACS, a real-time dataset crafted for Automatic Home Electrical Appliance Control Systems. This dynamic dataset originates from sensors embedded in an IoT framework, utilizing MQTT as its communication protocol. Departing from conventional cloud-processed datasets designed for big data, AHEACS adopts a sensor hub methodology, enabling seamless integration with machine learning in an IoT edge network for real-time analysis. The dataset comprises data from diverse sensors, including smart sockets linked to home electrical appliances, smart power meter, temperature and humidity sensors, and motion detectors sensor. AHEACS captures the on/off status of electrical appliances, total household power consumption, and environmental data such as temperature, humidity, and room occupancy. This paper investigates correlations among variables, specifically the air conditioner's on/off status concerning other factors, and devises a home automation model for air conditioners based on household behavior power consumption patterns and home environmental conditions. The dataset is minutely recorded over one week, amassing 20,403 data points, approximately equivalent to 2 weeks and 4 hours, stored in .csv format on a Raspberry Pi 4 functioning as an edge device. Analysis of the AHEACS dataset employs Python in the Google Colab environment, incorporating Exploratory Data Analysis (EDA) and six machine learning algorithms: ANN, DT, RF, SVM, LR, and XGBoost. Results from a case study focusing on the Air Conditioner's on/off status reveal a robust correlation with variables TP and Pb. The highest-performing algorithm, ANN, demonstrates an accuracy of 86.8%, MSE of 0.097, RMSE of 0.312, and R2 of 0.607.