Smart Journey in Istanbul: A Mobile Application in Smart Cities for Traffic Estimation by Harnessing Time Series

In recent decades, mobile applications (apps) have gained enormous popularity. Smart services for smart cities increasingly gain attention. The main goal of this proposed research was to present a new artificial intelligence (AI)powered mobile app on Istanbul’s traffic congestion forecast using traffic density data. It addresses the research question using time series approaches (long short-term memory (LSTM), Transformer, and eXtreme Gradient Boosting (XGBoost)) based on past data over the traffic load dataset combined with meteorological conditions. While previous studies were limited to direct Istanbul traffic forecasting, in this study, we focused on district-based traffic forecasting that can be queried with a mobile app. The proposed pipeline was tested on the summarized Istanbul traffic dataset for 6 main distinct districts (Fatih, Buyukcekmece, Atasehir, Kagithane, Tuzla, and Bagcilar). Analysis of the simulation results on predicted models will be discussed according to performance indicators such as mean absolute percentage error (MAPE), mean average error (MAE), and root mean squared error (RMSE). And then, it was observed that the Transformer model made the most accurate traffic prediction with a minimum MAE score for each district. The developed traffic forecasting prototype is expected to be a starting point for future products for a mobile app suitable for citizens’ daily use.