This paper presents a machine-learning-based approach for short-term forecasting of household electricity consumption. The study aims to model temporal consumption patterns and support intelligent energy management in residential environments. Historical power consumption data were collected, cleaned, normalized and transformed into supervised learning sequences using sliding window techniques. A Long Short-Term Memory (LSTM) neural network was developed to capture time-dependent characteristics of electricity usage. The model was trained using the Adam optimization algorithm and evaluated using standard regression metrics, including Mean Absolute Error (MAE), which indicated high prediction accuracy and robustness. To ensure practical applicability, the proposed system integrates edge computing principles. Experimental results demonstrate that deep learning-based time-series forecasting can effectively predict short-term energy consumption. The proposed approach contributes to smart home energy monitoring by providing a scalable, efficient and reliable solution, and supports sustainable electricity usage through data-driven decision-making. The findings highlight the importance of integrating predictive analytics into future intelligent energy systems.