Machine Learning Models and Mathematical Approaches for Predictive IoT Smart Parking

Area of Science:

• Computer Science
• Artificial Intelligence
• Internet of Things

Background:

• Smart parking systems are crucial for urban mobility.
• Accurate parking availability prediction is a key challenge.
• Existing systems often lack predictive accuracy.

Purpose of the Study:

• To develop an innovative approach for improving IoT-based smart parking systems.
• To enhance parking availability prediction accuracy using ML and AI.
• To integrate mathematical and autoregressive modeling strategies.

Main Methods:

• Developed and compared three regression-based ML models: random forest, gradient boosting, and LightGBM.
• Utilized autoregressive modeling with lagged features and Z-score normalization for time series forecasting.
• Employed Bayesian optimization for efficient hyperparameter tuning to minimize Root Mean Square Error (RMSE).

Main Results:

• The LightGBM model with lagged features demonstrated superior performance, achieving an R² of 0.9742 and an RMSE of 0.1580.
• Lagged features effectively captured temporal dependencies, outperforming other models.
• An IoT-based system architecture for real-time data collection was successfully developed and deployed.

Conclusions:

• The integration of ML, AI, and IoT significantly improves smart parking system efficiency.
• The proposed approach offers a scalable solution for urban mobility challenges.
• Accurate parking predictions contribute to reduced traffic congestion and enhanced city living.