ABSTRACT
The research focuses on the optimization of smart home systems using machine learning approaches. It is designed to help increase the efficiency and intelligence of home automation. The smart home environment includes a wide range of diverse sensors such as temperature, humidity, motion, as well as gas sensors installed in the environment. Furthermore, several actuating devices are used to ensure comfort, safety, and resource consumption. The actuating devices are integrated with a machine learning model that identifies the changes in the sensors’ signal and responds to them. Thus, actuating devices change their states to provide the home residents with the best environment. The performance of the model has been evaluated using the dataset comprised of 3200 records of sensor reading and the corresponding response of the actuating device. Machine learning models were developed and trained from learning instances to make predictions from sensor readings. The obtained results show that the Artificial Neural Networks are the most effective models. The evaluation demonstrates that the ANNs have high precision, recall, F1 score, and accuracy, whereas the loss was also low. Therefore, the ANNs proposed for predicting the actuating device response demonstrated high performance, which is critical in smart home optimization. The use of machine learning in smart home optimization has significant potential in monitoring and predictive maintenance in real-time conditions, as well as in energy and resource management. Such opportunities offered by advanced machine learning technology provide high compatibility and real-time performance.
