Abstract: To improve the precision of short term traffic flow prediction, this paper firstly analyzed the spatial-temporal characteristics and chaos of traffic flow data measured on Beijing Second Ring Road. Then on this basis, the chaos theory method was used to reconstruct the phase space of traffic flow time series, and an improved BP neural network model was proposed by employing the mind evolution algorithm. Finally, the reconstructed time series data were used as model input for traffic flow prediction. The results indicated that compared with the traditional BPNN model, the proposed model could reduce mean square error, mean absolute error and mean average percentage error by 31.11%，20.71% and 37.28% respectively, which proved that the proposed prediction model had more accurate prediction ability.

Key words: short term traffic flow prediction, chaos theory, phase space reconstruction, neural network, mind evolution algorithm