Predicting Household Electricity Consumption Using Machine Learning and Big Data Analytics

The growing dependence on natural resources and the progressively increasing demand for energy underscore the need of robust energy prediction capabilities to efficiently balance between supply and demand. Notably, residential building holds approximately 30–40% of total energy consumption, highlighting the crucial urge for accurate energy prediction capabilities. In this study, we propose a methodology for predicting energy consumption in residential buildings. The proposed methodology is structured into five key layers: 1. dataset composition, 2. time-based feature selection, 3. Apache Spark MLlib integration, 3. machine learning based model implementation, and 5. performance evaluation. This study emphasizes on mitigating the large-scale data processing using Apache Spark, a distributed processing framework, with outstanding scalability, robustness and flexibility in handling large-scale data analytics. For the validity of the proposed approach, we have used renowned evaluation metrics including Area Under the Curve (AUC), precision, recall, accuracy and F1-score. The result demonstrates that big data analytics significantly enhanced speed, computational complexity, robustness and prediction capabilities of ML models.