The increase in energy demand for buildings due to the development of technology and population growth requires accurate forecasting methods that satisfy the needs of society and at the same time reduce environmental effects, ensuring sustainable development in urban areas. Factors such as device density and geographic location significantly affect power consumption, which highlights the complexity of forecasting tasks. Various methods are used to predict future consumption patterns, including regression models, time-dependent models, and machine learning techniques. This paper reviews how Artificial Intelligence can be utilized to solve intricate nonlinear problems. Specifically, this paper discusses algorithms that enhance the efficiency of machine learning models, particularly for variable estimation. The packet algorithm is designed to make the Artificial Neural Networks method more efficient when dealing with heterogeneous data that displays temporal variations, such as those present in various types of months and days. In addition, a bagging method is introduced to improve the accuracy of neural networks. The Support Vector Machine model is a top performer in terms of performance. The voting Regression method and Decision Tree method produced more precise and accurate energy consumption predictions representing half of the deviation of the Artificial Neural Networks model.