Optimizing IoT-Driven Smart Cities with the Dynamic Leader Sibha Algorithm: A Novel Approach to Feature Selection and Hyperparameter Tuning

The rapid growth of Internet of Things (IoT) technologies has transformed modern urban environments into complex smart cities, generating vast amounts of high-dimensional, heterogeneous data. Effectively analyzing this data is crucial for optimizing urban infrastructure, enhancing quality of life, and supporting sustainable development. However, smart city data presents significant challenges, including non-linear dependencies, noisy signals, and high dimensionality. To address these challenges, this study proposes the Dynamic Leader Sibha Algorithm (DLSA), a novel metaheuristic optimization technique inspired by the structured counting dynamics of the Sibha. The DLSA was applied to the Smart Cities Index dataset, leveraging copula functions to model complex, multivariate dependencies and enhance predictive accuracy. The baseline machine learning (ML) evaluation revealed that the ExtraTreesRegressor achieved the lowest mean squared error (MSE) of 0.007462409, highlighting its superior initial performance. Following feature selection using the binary Dynamic Leader Sibha Algorithm (bSiba), the average error was reduced to 0.373245769, significantly improving data quality and model efficiency. Subsequent ML evaluation after feature selection further reduced the MSE of the ExtraTreesRegressor to 0.00151927, reflecting the effectiveness of dimensionality reduction. Finally, hyperparameter optimization using the DLSA achieved a remarkable MSE of $1.32249\times {10}^{-6}$ with the Siba + ExtraTreesRegressor combination, demonstrating the algorithm’s powerful optimization capabilities. These findings indicate that the DLSA framework can significantly enhance the predictive performance of IoT-driven smart city models, offering valuable insights for urban planners, policymakers, and technology developers seeking to build smarter, more resilient cities.