Institute of Transportation, MOTC

　　In this study, Lyapunov exponents and cellular automaton (CA) are applied to examine whether the short-term traffic flows on urban streets display stochastic or chaotic phenomena. Empirical tests were conducted out of five, ten and fifteen minutes’ traffic flow rates that are observed from some selected roadways in Taipei. We found strong evidence of nonlinear dependence in the short-term traffic flow dynamics and that chaotic time series model can successfully explain such nonlinear structures. In comparison with the conventional integrated autoregressive-moving average model (ARIMA), we also found that the chaos time series model obviously outperforms on the prediction accuracy of such short-term traffic flow dynamics.