Allocation Problems in Ride-Sharing Platforms: Online Matching With Offline Reusable Resources

Authors: John Dickerson, Karthik Sankararaman, Aravind Srinivasan, Pan Xu

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through a data-driven analysis on a massive openly-available dataset, we show our model is robust enough to capture the application of taxi dispatching services and ridesharing systems. We also present heuristics that perform well in practice.
Researcher Affiliation Academia John P. Dickerson University of Maryland, College Park, USA john@cs.umd.edu, Karthik A. Sankararaman University of Maryland, College Park, USA kabinav@cs.umd.edu, Aravind Srinivasan University of Maryland, College Park, USA srin@cs.umd.edu, University of Maryland, College Park, USA panxu@cs.umd.edu
Pseudocode Yes Algorithm 1: Simulation-based adaptive algorithm (ADAP(γ))
Open Source Code No The paper discusses an 'openly-available dataset' but does not provide any statement or link for the source code of their proposed methodology or algorithms.
Open Datasets Yes To validate the approaches presented in this paper, we use the New York City yellow cabs dataset,5 which contains the trip records for trips in Manhattan, Brooklyn, and Queens for the year 2013. 5http://www.andresmh.com/nyctaxitrips/
Dataset Splits No The paper mentions splitting data for 'training purposes' and 'testing purposes' but does not specify a separate 'validation' split for model tuning.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory).
Software Dependencies No The paper does not provide specific software dependencies with version numbers used for the experiments.
Experiment Setup Yes We set we = max(L1 αL2, 0), where α is a parameter capturing the subtle balance between the positive contribution from the trip distance and negative contribution from the docking distance to the final profit. We set α = 0.5 for the experiments. We consider each single day as the time horizon and set the total number of rounds T = 24 * 60 / 5 = 288 by discretizing the 24-hour period into a time-step of 5 minutes. In our algorithm, we optimized the value of ϵ and set it to ϵ = 0.1.