Daily Energy Usage Estimation Using Regression Based Ensemble Methods

Providing uninterrupted supply energy flow is of in-valuable importance for today's humanity. Because almost every device used today need constant energy to operate flawlessly. However, correct and efficient use of energy is a point that humanity should pay attention to and is often overlooked. Thanks to the developments in technology, the problem of efficient use of energy, in other words, the prediction of energy usage, can now be transferred to computers, more specifically, machine learning methods. From this point of view, in this study, energy usage prediction was considered as a problem that can be solved with machine learning methods. Specifically, the problem was addressed as a regression problem and two powerful ensemble methods, namely Random Forest Regressor (RFR) and Extra Tree Regressor (ETR) methods were used to solve the problem. Both methods were evaluated from the perspective of estimator and maximum depth numbers. Energy data collected from households in the Sceaux region of France for 4 years were used in experiments. Results suggested that mentioned methods were capable of successfully estimating energy usage.