A sustainable system for predicting appliance energy consumption based on machine learning

Nowadays, energy conservation is a top priority for sustainable societies, which focus on environmental and economic sustainability to address fossil fuel scarcity, climate change, and increasing environmental pollution. Therefore, planning for energy management and forecasting energy consumption is essential, especially with societies striving to achieve sustainable development. Predictive distribution plans for consumers and utilities can be improved using data mining-based models, the most common of which are big data-based Machine Learning (ML) models. This work presents a comprehensive ML model that combines the use of MATLAB for data reduction using Principal Component Analysis (PCA), and one of the latest data mining tools Orange 3 to build a classification model consisting of four basic classifiers: AdaBoost, Logistic Regression (LR), Naive Bayes (NB), and Stochastic Gradient Descent (SGD). The model relies on a dataset of energy consumption of different devices according to the Kaggle platform, where the energy consumption data was collected every 10 min and over approximately 4.5 months using m-bus energy meters. The model was tested based on the confusion matrix, and the results showed that AdaBoost outperformed other models in predicting energy consumption, with 100 % accuracy. In addition, LR, NB, and SGD had classification accuracy of 99.8 %, 99.7 %, and 99.4 %, respectively.