Appliances Energy Prediction Using Random Forest Classifier

Abstract

In this era of technological advancement, each and every appliances utilize electricity as the prime source of energy. In this scenario, it becomes important to predict the demand of electricity and thus tuning the operating parameters of the power plant for maximum utilization. In this paper, regression models for appliance energy consumption are trained, tested, and compared using various regression techniques. Prediction models used are data driven. Three various methods are used for training and testing of regression models: (1) Random forest (2) Artificial neural network with multilayer perceptron (3) Support vector machine with radial based function. Also, three data subsets are created from the original dataset and used for regression modeling. Comparisons of the above three regression methods are done by evaluating them with various statistical parameters. In this case, random forest method gives far better results compared to other two methods. It was able to explain 98.1% of correlation in training set and 98.05% of correlation for testing set.