An Empirical Study of With-in Day OD Prediction Using Taxi GPS Data in Singapore

The real-time OD prediction is a vital issue in DTA based traffic prediction systems. Previous researches along this direction are very sparse due to the low availability of OD volume observations. This study, utilizing the taxi GPS data collected in Singapore, demonstrated the effectiveness of different statistical models in predicting the future OD. The performance of four different classical statistical methods including historical average, ARIMA model, KNN method and ANN model are tested and compared using the dataset. The study has demonstrated that ANN models have highest overall prediction accuracy compared with other methods and can provide reliable prediction up to several hours range.