Artificial intelligence for parking forecasting: an extensive survey of machine learning techniques

Abstract

To address the parking challenges, this survey delves into the significant impact of machine learning (ML) on parking availability (PA) predictions. With swelling urban populations, efficient parking management has become paramount. PA prediction offers accurate, context-sensitive solutions for dynamic on-street and off-road parking scenarios, thereby promoting urban mobility and parking efficiency. However, traditional ML models, while contributory, struggled to capture complex contextual nuances and dependencies for effective predictions. The rapid advancements of deep learning offer promising avenues for sophisticated prediction models. This survey covers a wide spectrum, from PA definitions and relevant datasets to ML modules, features considered, and evaluation metrics. Additionally, the current limitations and future directions are also explored. This comprehensive review underscores the present contributions of ML in parking predictions and paves the way for refining and devising future developments to tackle the persistent parking issues.