PREDICTIVE MODELING OF HOUSEHOLD ENERGY CONSUMPTION IN THE USA: THE ROLE OF MACHINE LEARNING AND SOCIOECONOMIC FACTORS

Understanding the pattern of energy use at the household level becomes ever more urgent in light of growing concerns about climate change and resource sustainability in the USA. Energy use depends upon various factors, such as climate, household characteristics, and behavior. Of these, income, education, and size of the family are very vital socio-economic factors that depict energy consumption levels and their pattern. The utmost objective of this research project was to develop predictive models using machine learning techniques to analyze household energy consumption trends in the USA, integrating socioeconomic factors such as income, family size, and education. The dataset retrieved from Kaggle integrates detailed weather patterns with energy consumption data, putting into perspective the interaction between climatic variables and household energy use. It included key features such as temperature, humidity, wind speed, and precipitation, along with time-series data on energy consumption metrics like electricity and natural gas usage at the household level. It provided information on several geographic zones across extended periods, so seasonality and regional variations may be studied. It was complemented with metadata that included timestamps, energy pricing, and household attributes and should therefore be a rich resource for predictive modeling and extracting relationships between weather conditions and energy demand. For this research project, three models were selected: Logistic Regression, Random Forest, and Support Vector Machines, each possessing particular strengths for the nature of the problem. This study employed key performance metrics such as precision, recall, F1-Score, and accuracy. The Random Forest model had the highest value for accuracy, similarly, the highest AUC was for the Random Forest with the best AUC. As such, it was concluded that the Random Forest model provided the best trade-off between true positive rate and false positive rate and can be relied upon for this classification task. The machine learning models generate valuable predictions about household energy use. Particularly, Random Forest models, which are trained on socioeconomic and weather data to predict the likelihood of a given household having high energy usage. The predictions by such models can be used to help energy providers determine when to invoke tiered pricing or encourage energy-saving behavior.