Advanced Energy Consumption Prediction for Smart Cities Using Meta-Learning Layers

Predicting energy consumption in smart cities possesses challenges such as data heterogeneity, scalability, and model inefficiency. This study introduces an innovative meta-learning architecture that integrates advanced feature engineering to tackle these challenges. The model analyses datasets obtained from Kaggle, attaining outstanding metrics: MAE of 0.028, MSE of 0.002, AUC-ROC of 0.985, accuracy of 96.8%, and precision of 97.2%. The suggested method significantly outperforms baseline models (Linear Regression, Random Forest, LSTM). Significant advancements encompass the integration of a meta-learning layer for adaptive learning and the implementation of robust feature engineering for energy data. These developments provide a scalable and economical method for optimising energy in smart cities. The results support its excellence, facilitating sustainable urban planning and promoting smart city growth. Future study intends to improve computing efficiency and investigate more extensive datasets to increase applicability.