Reliable Machine Learning Models for Energy Optimization in Smart Green Cities

This dataset supports the manuscript entitled “Reliable Machine Learning Models for Energy Optimization in Smart Green Cities” submitted to the International Transactions on Artificial Intelligence (ITALIC), Vol. 4, No. 2, 2026. The study evaluates the reliability and predictive performance of machine learning models for urban energy demand forecasting and energy optimization in smart green city environments. The dataset contains reconstructed and structured research data based on the experimental design described in the manuscript. It includes hourly smart city energy consumption data, preprocessing attributes, model evaluation results, repeated execution outcomes, reliability statistics, interpretability consistency, energy optimization impact, and governance alignment indicators. The dataset is designed to support transparency, reproducibility, and further academic analysis related to AI-based energy management. The research uses 17,520 hourly observations representing energy consumption patterns from January 2022 to December 2023. The dataset includes energy consumption as the target variable, along with weather-related and temporal features such as temperature, humidity, hour of day, day of week, weekday/weekend status, and rolling mean consumption. The data structure follows the chronological train-test partition described in the study, consisting of 14,016 training observations and 3,504 testing observations. The machine learning models evaluated in this dataset include Random Forest, Support Vector Machine, and Long Short-Term Memory (LSTM). Each model was tested across 30 independent execution runs using performance metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), coefficient of determination (R²), reliability score, and interpretability consistency. The dataset also includes summarized findings showing that the LSTM model achieved the best performance, with MAE of 0.31, RMSE of 0.45, R² of 0.93, and a reliability score of 0.912. This dataset can be used for research replication, comparative machine learning analysis, reliability-oriented AI evaluation, smart city energy forecasting, and sustainable energy management studies. It also supports the objectives of SDG 7 Affordable and Clean Energy, SDG 11 Sustainable Cities and Communities, and SDG 13 Climate Action, as discussed in the manuscript.