Hybrid parking space prediction model: integrating arima, lstm, and bpnn for smart city development
Abstract
Introduction: Parking space prediction is essential for addressing traffic congestion challenges and low parking availability in urban areas. It is a significant aspect of smart city development and a sustainable environment. The present research mainly focuses on predicting parking space using smart devices that collect time-series data, which is complex and unpredictable. Smart cities require an efficient parking prediction system. Methods: We present a smart parking space prediction model that combines Auto-Regressive Integrated Moving Average (ARIMA) and Long Short-Term Memory (LSTM) models. The novelty of the proposed model is that it is hybridized using a back propagation neural network (BPNN). In contrast to the conventional assumption of a linear relationship between the predicted results of ARIMA and LSTM models, this approach employs BPNN to uncover the unidentified objective and establish the connection among the predicted values. The model utilizes the ARIMA model for handling linear values and the LSTM model for handling non-linear values of the (Internet of Things) IoT dataset. Results: The "Melbourne" dataset is used for evaluation. The proposed hybrid model achieves the minimum MSE, MAE, and RMSE values of 0.32, 0.48, and 0.56, respectively. Conclusion: This model potentially improves parking space prediction, contributes to sustainable and economically smart cities, and enhances citizens' quality of life.