A Meta-Analysis of AI-Driven Geospatial Analytics for Predictive Maintenance of Critical Infrastructure in Developing Economies

Abstract

This study conducted a quantitative meta-analysis to evaluate the effectiveness of AI-driven geospatial analytics in predictive maintenance of critical infrastructure within developing economies. The research synthesized data from 52 empirical studies encompassing 18,745 infrastructure units across transportation, energy, water systems, and telecommunications sectors. The analysis aimed to assess improvements in predictive accuracy, maintenance efficiency, cost reduction, and system reliability achieved through the integration of artificial intelligence and geospatial data. The findings indicated that AI models incorporating spatial-temporal data significantly outperformed non-geospatial models, achieving a mean prediction accuracy of 91.2% compared to 78.8%, reflecting an average improvement of 12.4%. Regression-based models demonstrated an error reduction of 18.7%, while classification models improved failure detection accuracy by 14.2%. The overall pooled effect size was 0.68, indicating a moderate to strong impact on predictive maintenance performance. Sectoral analysis revealed that transportation and energy systems achieved higher predictive accuracies of 92.4% and 90.8%, respectively, compared to water systems at 86.3% and telecommunications at 84.7%. The results further showed that predictive maintenance approaches contributed to a 24.5% reduction in maintenance costs and a 19.2% decrease in system downtime. Subgroup analysis confirmed that deep learning and ensemble methods achieved higher effect sizes, ranging from 0.72 to 0.77, particularly when applied to integrated datasets combining satellite imagery and sensor data, which achieved predictive accuracy up to 93.3%. Additionally, 84.6% of the included studies reported statistically significant results, reinforcing the robustness of the findings. The study highlighted the critical role of data integration, algorithm selection, and contextual factors such as data availability and infrastructure type in influencing model performance. Overall, the results demonstrated that AI-driven geospatial analytics provides a highly effective and scalable approach for enhancing predictive maintenance strategies in infrastructure systems, particularly within resource-constrained environments.