Developing a Machine Learning Model for a Smart Home System

  Machine learning improves the efficiency of smart home systems, allows home automation systems to regulate heating and cooling systems, lighting, room temperature and other parameters. Data variability and improvement of such systems require constant expansion of data sets, retraining or additional training of machine learning models, modification of algorithms and architectures. The article presents models for predicting heating and cooling loads of a house based on machine learning methods. The results of exploratory data analysis, construction of regression models for predicting the load of heating and cooling systems are presented. The efficiency of selecting hyperparameter values based on the grid search method is shown. A neural network model is considered that allows simultaneous prediction of the load of heating and cooling systems. The accuracy is assessed and the models are compared based on regression quality metrics.