| dc.description.abstract | Now, the network of freeway is being completed. In presupposition that forecasting results are exact. The travel time forecasting information not only can let passengers directly realize the situation of traffic flow they will in, but also can let the traffic management make a suitable decision to fit this situation and decide a proper route of passenger’s needs to make the best use of the network of freeway in the future.
Because that our link of freeway hasn’t use Automatic Vehicle Identification system to collect travel time data, so this study focus on the weekday’s situation of traffic flow, using simulation to gain travel time data and considering all kinds of related influential factors, such as the updating time period of information, distance between two neighborly collecting stations, and AVI rate. Furthermore, Using four forecasting models, such as single exponential smoothing method, Holt’s exponential smoothing method, autoregressive integrated moving average method, and back-propagation network, to test and analyze if forecasting results are exact. Hoping to offer practicable and exact travel time forecasting information, and using it as the basis of how passengers choose their routes.
From the result of all kinds of test, we can know that it can have better forecast effect when information update takes five minutes, keeps one kilometer between two neighborly collecting stations. Besides, in four travel time forecasting models. HES forecast effect is the worst, and other’s forecast effects are almost the same but we still can have better forecast effect much easily from BPN forecasting model. Lastly, we can regard the result of all kinds of test as reference which is related to the traffic management. | en_US |