Analysis and modeling of large-scale systems : taxis and social polling
- Technological advances in transportation have fundamentally revolutionized the way people live and move. The foundational blocks of technology are: (i) cheap and fine-grained sensors of movement, (ii) a ubiquitous communication network for collecting data from the sensors, (iii) a cloud-based platform for storing, processing and analyzing the movement data, and (iv) a high penetration of smart phones over which to offer novel transportation-based services. Thus, in a short span of time, the way people commute in public transit or hire car systems has changed dramatically; the way packages and food is delivered is changing; and even the travel and leisure industry is being disrupted. These massive societal trends have also generated many interesting technical challenges, notably in terms of "a big data system for things that move". In this thesis, we describe the challenges of collecting and processing real-world movement data from a large-scale network of vehicles and passengers. We then describe a big data system and algorithms for healing noisy real-world data and then processing, analyzing and visualizing it. While we have made use of such a system for our analysis, our key contribution is a methodology for modeling large-scale movement processes based on data. Using these models we are able to answer questions such as: What factors affect the operation of a large-scale transportation network? What factors limit the efficiency of a large-scale transportation network and how can the efficiency be improved? We consider the above questions in the context of data from New York City yellow taxicab rides for 2013, which has over 170 million trips from a fleet of more than 14,000 taxis. From an analysis of the data we obtain, in addition to insights about the operation of the taxi fleet, the following: (i) correlations between taxi operation and cultural activities in the city, (ii) economic reasons for shift change times, (iii) the impact of severe weather on travel in the city, and (iv) the impact of congestion on taxi trips. Using our modeling framework, we consider methods of enhancing the efficiency of the taxi system in terms of reducing idle time for the taxis and, hence, using a smaller number of vehicles for handling the same number of rides. We obtain a network-flow based model for reassigning trips to taxis which reduces the number of taxis required to fulfill all trips in a city. One finding of our model is that only 72% of yellow taxicabs are necessary to finish all observed trips in New York City, following the same schedule of pickups and drop-offs. Furthermore, after investigating trips reassigned by our model, we discover patterns which provide useful insights into possible directions of improvement for the current taxi system.
|Type of resource
|electronic; electronic resource; remote
|1 online resource.
|Stanford University, Department of Computer Science.
|Prabhakar, Balaji, 1967-
|Prabhakar, Balaji, 1967-
|Statement of responsibility
|Submitted to the Department of Computer Science.
|Thesis (Ph.D.)--Stanford University, 2015.
- © 2015 by Chenguang Zhu
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...