Artificial Neural Networks in Smart City Service Delivery: A Systematic Review of Architectures, Evidence Quality, and the Missing Link to Urban Economic Development

Abstract

Although smart city investments have surged in fields such as healthcare, infrastructure and waste management, very little is known about the quality of evidence or its association with urban economic development. This systematic review, based on PRISMA 2020 guidelines, searched four major academic databases for peer-reviewed literature published between January 2011 and April 2025. 58 included studies were appraised independently by two reviewers using MMAT 2018 (mean score 76.5%, Cohen’s κ 0.738) and were divided into four thematic clusters: Healthcare AI (n = 14), Smart Infrastructure (n = 12), Waste Management AI (n = 11) and Smart Cities—Saudi/GCC (n = 21). CNNs dominate the image-based waste classification; LSTM and ensemble methods each outperform in different time-series and tabular waste forecasting tasks. Quantified benefits include a 0.773 AUC-ROC across 78 disease prediction tasks, a 16% decrease in travel time, and a 35% decrease in waste collection frequency. Direct empirical linkage between the service efficiency generated by the ANN and economic development at the zone level remains largely unestablished in the reviewed literature-an issue that is particularly salient in the GCC context. The dual-stage ANN mediation model is suggested as the main research priority in this review.