Estimating demand for paratransit bus services for people with disability

公共交通机构有责任提供高效、高效和可靠的服务，让所有市民，特别是严重依赖公共交通服务的残疾乘客，都能对服务产生积极的印象。公共交通机构及其残疾乘客面临的一个日益严峻的挑战是预测对辅助交通和按需服务的需求。辅助交通服务是一种特殊的服务，残疾乘客可以在一天中指定的时间要求直接从家中乘车到他们想要的目的地。在多伦多这样一个拥有270万居民的城市，多伦多交通委员会（TTC）每天会收到大约1.3万份关于辅助交通和按需服务的请求。因此，TTC必须每天制定时刻表，以适应所有的辅助交通和按需需求。作为回应，像Giro这样的公司开发了软件来帮助公共交通机构准备这些每日时间表，包括分配按需出租车服务，以弥补没有公共汽车可用时的空白。由Giro开发的哈斯特斯-随需应变软件使用一系列优化算法工作，这些算法依赖于一组假设，可以快速生成时刻表，供公共交通机构立即使用。哈斯特斯-随需应变软件面临的一个主要挑战是预测一天或一小时的需求。缺乏这样的需求信息会增加计算时间，导致调度延迟，并可能导致依赖需求假设来达到调度问题的最佳解决方案，而这可能由于缺乏信息而不正确。在这个项目中，我将使用TTC 2年的存档数据，当时他们使用Hastus-OnDemand软件进行预订，并将与Giro合作，为Hastus-OnDemand软件添加需求预测模块。这些数据目前归Giro所有，并根据一项研发协议提供访问权限。作为项目的一部分，将开发一个通用需求模型，使Giro能够将需求模型作为哈斯图斯-随需应变软件的输入，以帮助预测使用哈斯图斯-随需应变软件的其他地区的预期每日需求。

Public transport agencies have the responsibility of providing efficient, productive, and reliable service that is perceived positively by all members of the public, especially for passengers with disabilities who are heavily reliant on this service. One growing challenge for public transport agencies and their passengers with disabilities is predicting the demand for paratransit and on-demand services. Paratransit service is a special service where passengers with disabilities may request a ride directly from their home to their desired destination at a time they specify in the day. In a city with 2.7 million residents like Toronto, the Toronto Transit Commission (TTC) receives around 13,000 daily requests for paratransit and on-demand services. As a result, the TTC must develop schedules on a daily basis to accommodate all paratransit and on-demand requests. In response, companies like Giro has developed software to help public transport agencies prepare these daily schedules, including allocating on-demand taxi services to cover the gap when no paratransit bus is available. The Hastus-OnDemand software developed by Giro works using a series of optimization algorithms that depend on a set of assumptions made to enable quick schedule generation for immediate use by public transport agencies. A major challenge that the Hastus-OnDemand software faces is the prediction of demand in a day or by the hour. The lack of such demand information can add to computation time, causing delays in scheduling, and can cause a reliance on demand assumptions to reach an optimum solution for the scheduling problem that may not be correct due to the lack of information.In this project, I will be using 2 years of archived data from the TTC when they used the Hastus-OnDemand software to do bookings and will work with Giro to add a demand prediction module to the Hastus-OnDemand software. The data is currently owned by Giro and access is provided under an agreement for research and development. A generic demand model will be developed as part of the project that will enable Giro to include a demand model as an input in the Hastus-OnDemand software to help in predicting expected daily demand in other regions where the Hastus-OnDemand software is used.