Prediction of traffic flow time series data on Jakarta-Cikampek toll road using a chaotic approach and local linear approximation method

The Jakarta–Cikampek Toll Road is widely recognised as a critical corridor linking Jakarta with major industrial centres and expanding residential areas in West Java. Traffic along this route is frequently dense and exhibits noticeable fluctuations over time. At certain periods, these variations reveal nonlinear patterns shaped not only by commuter movements and freight transport but also by broader economic activity. This study focuses on short-term traffic forecasting by modelling traffic flow as a nonlinear dynamical system rather than a purely stochastic process. A chaos-based framework is applied to the traffic flow time series through two main stages: first, the presence of chaotic behaviour is examined using the 0–1 test; second, short-term forecasting is performed using the Local Linear Approximation Method (LLAM), which utilises neighbouring trajectories in phase space. Hourly traffic flow data recorded over seven consecutive days are analysed, with 144 observations used for training and 24 for testing. The 0–1 test confirms the existence of chaotic dynamics within the series. During the forecasting stage, LLAM demonstrates reasonable agreement with observed traffic flow, achieving a Pearson correlation coefficient (r) of 0.8964, a Mean Absolute Error (MAE) of 159.41 vehicles per hour, and a Root Mean Squared Error (RMSE) of 203.42 vehicles per hour. These findings indicate that chaos-based modelling provides a dependable approach for short-term traffic forecasting. Beyond predictive accuracy, the framework offers practical value by supporting quicker operational responses, improved congestion management, and more effective short-term planning in dynamic toll-road environments.