Predictive Modeling of Energy Consumption in U.S. Smart Homes Using IoT Sensor Data and Climate Variables

Abstract

This research addresses the challenge of monitoring and optimizing energy consumption in smart homes, due to the high volume of data generated by IoT devices and its relationship with climate variables. The central problem lies in the absence of an IoT system capable of processing, integrating, and transforming large volumes of data into useful information for energy decision-making. The objective was to design, implement, and evaluate an intelligent system based on IoT, AWS, DuckDB, Python, and Power BI for monitoring, analyzing, and visualizing energy consumption in smart homes. For this purpose, the agile Scrum methodology was used, due to its iterative and incremental approach. A dataset of 503,910 records was used, from which a 5% sample equivalent to 25,195 observations was selected. The developed architecture integrated IoT sensor data, AWS storage, data cleaning and loading through Python, queries with DuckDB, and interactive visualization in Power BI. The results showed improvements in query time, report generation, reduction of incorrect data, and elimination of missing data in the experimental group after the implementation of the system. In conclusion, the proposed architecture proved to be efficient and reliable for supporting timely decisions aimed at energy efficiency in smart homes.