Intelligent parking space management: a binary classification approach for detecting vacant spots
Hanae Errousso,El Arbi Abdellaoui Alaoui,Siham Benhadou,Anand Nayyar
DOI: https://doi.org/10.1007/s11042-024-18777-w
IF: 2.577
2024-03-23
Multimedia Tools and Applications
Abstract:In the wake of escalating urban vehicle proliferation, major cities grapple with intensified traffic congestion, elevated air pollution, and acute scarcity of parking spaces. Addressing this issue, the research paper proposes an advanced Intelligent Parking System (IPS) using artificial intelligence, transforming the traditional regression problem of parking space prediction into a binary classification task through a novel algorithm. This algorithm inserts a binary variable into existing databases to model each parking space's availability, thus adapting tabular parking history data ("data architecture" most widely adopted by intelligent parking solutions already in use) to a classification problem. Once data has been prepared and processed, parking spaces are classified according to their availability employing logistic regression with regularization techniques, achieving superior performance in precision, recall, and F-score metrics. Specifically, the proposed model demonstrates remarkable accuracy (up to 92.52% for Log-log with L1 &L2 regularization), high sensitivity (Sn: 96.90% for Probit with L1 &L2), and specificity (Sp: 94.42% for Log-log with L1 &L2), outperforming contemporary approaches. The comparative analysis also reveals the model's efficiency in terms of accuracy and specificity against recent works. Additionally, the evaluation of time and energy complexity underscores the practical applicability in real-world urban environments. Thus, the research contributes significantly to smart city development and sustainable urban transportation infrastructure, laying groundwork for future advancements.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering