王婧

       随着城市人口的日益膨胀，机动车拥有量及道路交通流量急剧增加，城市交通的压力越来越大，大部分国家将优先发展公共交通作为解决城市交通问题切实可行的措施，如何做好公交规划，保证公交系统正常、高效有序地运营任重道远。公交客流调查是公交规划过程中一项基础信息采集技术，其方法的优劣直接影响到所获取客流总量信息的质量，进而影响公交规划与管理工作的效果，因此非常有必要对公交客流总量的获取方法进行深入分析探讨，为政府对公交线路补贴提供理论依据。

       本文首先梳理常用公交客流信息采集方法，结合数据融合、聚类分析思想，提出人工调查与公交IC卡数据融合获取公交客流总量的方法，对公交IC卡数据进行聚类分析，从日变、月变、周变、时变四个角度阐释公交线路总体客流特征；其次，选择影响公交客流量变化的随机型因素，构建公交客流影响因素分析多元线性回归（MLR）模型，结合模型输出及客流特征分析结果得到公交客流量调查日期选取分调查月份、调查星期、调查日期三个步骤；再次，从调查成本和结果精度两个角度出发，借鉴最优化思想计算调查最佳抽样率，将一日小时客流量进行有序样本聚类，划分抽样调查应覆盖的客流时段；结合调查线路按车辆统计的详细调度信息，遵循“选中车辆在一天内的班次尽可能多”及“调查车辆一日内的运营班次需覆盖每一类客流时段”两个原则确定调查车辆，被选中车辆的首班次发车时间定为抽样调查开始时间；最后，介绍公交客流数据融合分析方法：按线路客流时段划分结果统计调查线路各时段IC卡刷卡比例，作为同类线路IC卡比例参考值，运行MySQL软件程序得到各线路客流总量。从管理者和乘客两个角度给出公交冷热线路评价指标，构建评价体系，得到各条线路冷热程度评价等级，从管理者角度出发，对线路补贴提出建议。在理论研究的基础上，以上海市奉贤区33条公交线路为例，对理论方法进行了实际应用。

关键词：公共交通，客流调查，人工调查，IC卡，数据融合

    With the expanding of urban population, the dramatically increasing of car ownership and road traffic, the pressure on urban traffic is also intensifying. Most countries treat transit priority as practical measures to solve urban traffic problems. There is a long way to go to make a good transit planning, which can ensur transit system operates normally, efficiently and orderly. Bus ridership survey is a basic information collection technology in the process of transit planning, the merits and demerits of the method will affect the quality of the obtained ridership information directly, and then affect the effectiveness of transit planning and management. Therefore, an in-depth analysis on methods of obtaining ridership information is of great necessary, which can provide a theoretical basis for the government to provide subsidies to bus routes.

    This paper combed the common bus ridership information collection methods. Firstly, based on the data fusion and cluster analysis, this research put forward a method of combining artificial investigation and bus IC card data to obtain bus ridership information. The method of clustering analysis was carried out on the bus IC card data, using monthly, weekly, daily, and hour varying ridership to expound the overall bus ridership characteristics; Secondly, I chose random factors that resulting in the variation of bus ridership to construct multiple linear regression (MLR) model. Through combining with the MLR model output and the ridership characteristics, we concluded that the selection of survey date required three steps to determine the survey: month, week and date respectively; Thirdly, in view of survey cost and precision, I used optimization ideas to calculate the optimum survey sampling rate, with ordered sample cluster method to analyze daily hour-ridership, in order to divide ridership time interval that should be coverd by survey period. Considering the dispatching information according to operational vehicles, the selection of survey vehicles should follow two principles: one is that“the selected vehicles’ flights should as many as possible in a day”, the other is that“the selected vehicles’ flights in the daytime should cover all kinds of ridership time intervals”, the first flight departure time of the selected vehicle will be the start time of sampling. Finally, I introduced the method of bus ridership data fusion. According to the division results of ridership time intervals for various routes, I caculated the ratio of passengers using IC card during various period, which will be treated as the reference basic value for extending the same route cluster, and then obtained the ridership information of all routes by running MySQL software program. I gave the evaluation index to build evaluation method which can be used to determine the cold or hot status of every route from the perspective of managers and passenger. Then the management can get some reasonal suggestions on route subsidies from the managers’ perspective. On the basis of theoretical research, I chose 33 transit routes in Fengxian district in Shanghai as an example to apply the theories and methods.

Key Words: transit, ridership survey, manual survey, IC card, data fusion