<p>The absence of real-time parking information is the primary cause of drivers’ endless quest for parking spots – an activity that leads to wasted time and fuel, and increased pollution. Researchers have addressed this issue by developing parking availability prediction systems that employ various modeling techniques; however, there is a lack of extensive studies on parking occupancy predictions and a limited number of analyses of a sparse collection of methodologies. This study aimed to surpass previous efforts in scope and synthesis by conducting a thorough analysis of 108 research papers, journals, and articles selected from an initial pool of 214, all published post-2010 to provide a detailed review of parking occupancy prediction methods. One of the key findings was the diverse methods in which data is collected and utilized: combining inputs from parking sensors, cameras, mobile crowdsourcing, and manual counts to enhance model accuracy and provide a holistic view of urban parking dynamics. The analysis underscores the importance of incorporating temporal, geospatial, socio-economic, and policy-related factors into models to reflect the multifaceted nature of parking occupancy. Hybrid models that combine different approaches were shown to improve predictions by capturing a wider range of patterns and relationships in parking occupancy dynamics. Neural network models, particularly convolutional and recurrent neural networks, were found to excel at processing complex spatial and temporal patterns. Time series models, such as autoregressive integrated moving average models, were deemed suitable for capturing temporal patterns, and long- and short-term memory models showed potential for predicting short-term parking availability. The findings of this study will aid researchers in identifying areas where further research is needed and will help city planners make informed decisions about parking infrastructure and policies.</p>