A review of research on urban parking prediction

Abstract

The rapid growth of urban traffic has intensified daily congestion, affecting both traffic flow and parking. Accurate parking prediction plays a vital role in effectively managing limited parking resources and is essential for the successful implementation of advanced intelligent systems. In an effort to comprehensively assess the latest developments in parking prediction, we curated a dataset of 639 articles spanning from 2010 to the present, using the Scopus database. Initially, we performed a bibliometric analysis utilizing VOSviewer software. These findings not only illuminate emerging trends within the parking prediction field but also provide strategic guidance for its progression. Subsequently, we categorized advancements in three focal areas: behavior prediction, demand prediction, and parking space prediction. A comprehensive overview of the present research status and future directions was then provided. The findings underscore the substantial progress achieved in current parking prediction models, achieved through diverse avenues like multi-source data integration, multi-variable feature extraction, nonlinear relationship modeling, deep learning techniques application, and ensemble model utilization. These innovative endeavors have not only pushed the theoretical boundaries of parking prediction but also significantly heightened the precision and applicability of predictive models in practical scenarios. Prospective research should explore avenues such as processing unstructured parking datasets, developing predictive models for small-scale data, mitigating noise interference in parking data, and harnessing potent platform fusion techniques. This study's significance transcends guiding and catalyzing advancement in academic and practical domains; it holds paramount relevance across academic research, technological innovation, decision-making support, business applications, and policy formulation.