Identification Method Of Bus Transfer Study Based On IC Card And AVL Data

With continuous increase of urban construction scale,the city expands outward rapidly,and the distance travel of residents keeps increasing.However,conventional bus network cannot meet the direct bus service between every OD(Origin and Destination)pairs,so bus transfer cannot be avoided.With the development of intelligent transport system,it produces huge amounts of data which provides a possible to study transfer.Based on the inferred information of board and alight,transfer information can be identified according to the characteristics of time and space.Combining with the travel characteristics of transfer passengers,this paper proposes a two-stage method for identification of transfer,that is,preliminary identification and depth identification.Firstly,identify the intermediate transfer process in the continuous travel stage is transfer.Secondly,it continues to conduct in-depth identification in combination with the passenger travel chain on the basis of the preliminary identification result of transfer.First of all,in the preliminary identification of the transfer process,which uses the platform provided by the existing map to obtain the walking path distance between stations to replace the Euclidean distance making the identification process more reasonable and accurate.In addition,the identification rules are refined,such as considering different types of passengers walking speed,combined with the lights on the road to set redundant time,the bus in the site of the vehicle congestion,the chance to get in the car.At last,on the basis of the preliminary identification result,the depth identification is carried on to identify the short-term activities which are similar to the transfer.Taking the one-week travel data of Chengdu as an example,the result shows that the average transfer coefficient is 1.256.Then compare to the accuracy of base data between this article and previous studies to illustrate the precise of the recognition is improved,and it can increase about 20% accuracy in the route distance and the arrival time error calculated with this data can be reduced by 14.44% compared with Euclidean distance.And verify the results of the paper by combining with the index of transfer and short-term activity in the literature.Based on this,compare the results of the paper with the methods of predecessors used,such as fixed time interval recognition method,based on the principle of "to go" criterion and not considering space travel characteristics of chain determination method.And it analyzes the false negatives and miscalculation one by one.The results show that the congestion factor considered in this paper can reduce the missed sentences by 2.9%,and the transfer taking into account the pass-by travel can reduce the missed sentences by 8% compared with the principle of "to go" criterion.Compared with considering only the detour coefficient,the single-path and multi-path factors considered in this paper can reduce the misjudgment by 3.49%.Finally,the paper analyzes distribution of the transfer identification results in time and space,and it examines the spatial distribution of the station with a long waiting time for transfer,a crowded vehicle that cannot take the first target vehicle in time and a large transfer distance.