Machine Learning-Driven Evaluation of Sustainability Performance of Building Materials in Kenya

Abstract: Sustainability in Kenya’s construction sector faces critical challenges due to fragmented data systems, weak regulatory enforcement, and limited analytical frameworks for evaluating building material performance. Although sustainable construction has gained prominence, decision-making remains largely subjective and policy-driven rather than performance-evaluated. This study addresses the problem by developing a machine learning–driven framework to quantitatively assess the Sustainability Performance Index (SPI) of building materials in Kenya. Using survey data from 328 construction professionals and twenty normalized indicators spanning economic, environmental, social, and institutional dimensions, an Artificial Neural Network (ANN) model was trained and validated through 10-fold cross-validation. The ANN captured nonlinear dependencies among the SEET (Social, Environmental, Economic, Technological) parameters, achieving a predictive accuracy of and RMSE . Shapley Additive Explanations (SHAP) and Permutation Feature Importance analyses revealed Durability, Energy Efficiency, and Waste Reduction as dominant predictors, with Policy Enforcement and Awareness Level as critical institutional amplifiers. The findings demonstrate that the proposed ANN framework effectively operationalizes sustainability evaluation, providing policymakers and practitioners with an interpretable, evidence-based tool for material selection and performance benchmarking. The study concludes that integrating machine learning into sustainability assessment enhances transparency and adaptive governance in Kenya’s built environment. It recommends embedding such data-driven SPI models into national building codes, procurement systems, and housing policy audits to align construction practices with the Sustainable Development Goals (SDGs 9, 11, and 12). Future research should expand empirical datasets, integrate lifecycle costing, and hybridize neural networks with fuzzy or ensemble models to improve generalization across regional contexts. Keywords: Artificial Neural Network, Building Materials, Machine Learning, Sustainability, Sustainability Performance Index. Title: Machine Learning-Driven Evaluation of Sustainability Performance of Building Materials in Kenya Author: Edna Odongo Wayodi, Absalom H. V. Lamka, George Kinoti King’oriah International Journal of Novel Research in Civil Structural and Earth Sciences ISSN 2394-7357 Vol. 12, Issue 3, September 2025 - December 2025 Page No: 33-47 Novelty Journals Website: www.noveltyjournals.com Published Date: 08-October-2025 DOI: https://doi.org/10.5281/zenodo.17294308 Paper Download Link (Source) https://www.noveltyjournals.com/upload/paper/Machine%20Learning-Driven%20Evaluation-08102025-5.pdf