降雨量对OD 行程时间可靠性影响分析

为详细研究降雨量对OD行程时间可靠性的影响，基于Uber 出行共享的3 年美国波士顿10 对OD行程时间数据及WeatherUnderground 网站提供的小时历史天气，构建了OD行程时间高斯混合模型(GMM). 模型参数利用EM法进行求解，K 值根据K-S 检验后的P 值 (大于0.500 0)进行确定，模型分位数利用二分法进行求解. 提出一种基于缓冲指数(BI)的新指标——缓冲指数变化率(BIVR)作为定量评估指标. 结果表明：降雨会降低总体OD行程时间可靠性，降低效果随降雨量提高而增强，但增强效果并不明显；尽管可能性较低，但当降雨处于次要影响因素时可能提高可靠性；小雨天气可视为正常天气；雨天可靠性显著低于正常天气，居民在雨天(除小雨外)出行应预留更多时间.

Analysis of Effects of Rainfall on Origin-destinationbased Travel Time Reliability

With the purpose of investigating the effects of various rainfalls on the origin-destination(OD)-based travel time reliability (TTR) in detail, Gaussian mixture models of OD-based travel time were constructed based on three-year travel time data of ten OD pairs in Boston, USA, shared by Uber movement, and the hourly historical weather provided by WeatherUnderground website. And the model parameters, K -value and percentile-value were solved by expectation-maximum algorithm, the P-value (greater than 0.500 0) after K-S test and bisection method, respectively. Additionally, buffer index variation ratio (BIVR) is proposed for quantitative analysis. The results show that: rain can reduce the overall ODTTR, and the reduction effect increases with the rainfall increase, but the gaps are unobvious; Despite lower probability, ODTTR may be improved when rain as a secondary factor; Light rain can be regarded as normal weather; The ODTTR of rainy days is significantly lower than that of normal weather, so residents should reserve more time for travel in rainy days (except light rain).