基于出租车运营数据和POI数据的出行目的识别

为了有效获取出租车乘客出行目的，提出了一种基于出租车运营数据和POI(Point of Interest)数据的出行目的识别方法.构建了基于乘客出行特征和下车所属POI点类别的乘客出行目的识别模型，该方法从出行特征及乘客下车点最终可能到达的目的地所属POI点类型两个方面确定乘客的出行目的.为了验证所提方法的有效性及实用性，对成都地区展开了出租车出行调查，并利用调查数据对模型进行了精度验证.结果发现，相比于现有的利用出行特征推断出行目的的方法，本文提出的决策树+POI(II)能够提高最终识别准确度15.76%.最后，将所提方法应用于成都 1周的实际运营数据中，成功地识别出219 942名乘客的出行目的，说明该方法能够应用于实际数据量较大的出行目的识别.本文提出的方法，可以作为出行调查的辅助手段.

Trip-purpose-identification Based on Taxi Operating Data and POI Data

An effective trip-purpose-identification method is proposed based on taxi operating data and POI (point of interest) data to obtain the trip purpose of taxi passengers. A trip-purpose-identification model on the account of trip characteristic and type of drop-off points, which determines the destinations of taxi passengers from two aspects including trip characteristic and POI type of possible destinations. To verify the validity and practicability of the proposed method, an investigation into servicing taxi in Chengdu is launched, and the accuracy of the proposed model is verified by exploiting collected data. It turns out that the proposed method, decision-making tree plus POI(II), can improve the ultimate identification accuracy up to 15.76% compared with the existing method that deduces trip purpose upon trip characteristic. Finally, the proposed method is successfully applied to identify the trip purposes of actual one-week operation data, containing 219 942 taxi passengers, which implies this method can be applied into trip purpose identification with large-scale operation data. The proposed method can be used as an assistant method to trip survey.