Household Power Consumption Forecasting using IoT Smart Home Data


The use of the forecasting system is becoming more prominent in recent years. One of the implementations of the forecasting system is to predict electricity consumption demand. In this paper, we have developed a forecasting system for household electricity consumption using a well-known Extreme Gradient Boosting algorithm. We utilized time-series data from a smart meter dataset to make a predictive model. First, we evaluated the importance of time-series feature from the dataset and resampled the original dataset. Then, we used the resampled data to train the model and calculated training loss function. Our experimental studies with real IoT Smart Home data demonstrate that our forecasting system works well with small dataset using one-hour downsampling on the dataset.