Data fusion techniques are applied to traveler information services and used to build an accident duration estimation function. The accident duration is estimated at the initial occasion of an accident. The data fusion procedure clusters the values of each factor into a small number of intervals and effectively smoothes the data noise to the model. In most experiments, the mean absolute percentage errors of the estimated outputs are under 25%, indicating an acceptable forecasting effect. Through the factor sensitivity analysis, time of day, number of vehicles & vehicle type involved in accidents, and geometry have high significance in conducting the accident duration models. The results confirm that the data fusion techniques are practical and reliable for developing traveler information systems. This study is granted by National Science Council, Taiwan, under the project number NSC93-2218-E-006-094. It shows very promising practical applicability of the proposed models in the Intelligent Transportation Systems context.