Data Analysis and Optimization for (Citi)Bike Sharing

Authors: Eoin O'Mahony, David Shmoys

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We tested the integer programming approaches on real world instances gathered from actual system data. We implemented the IP in Gurobi (Gurobi Optimization 2014) and carried out a number of experiments. The experiments were run on Linux machines with 2 Intel x5690s running at 3.46GHZ, a 128GB SSD and 96GB RAM.
Researcher Affiliation Academia Cornell University Department of Computer Science1 School of Operations Research and Information Engineering2
Pseudocode No The paper provides mathematical formulations and descriptions of algorithms (e.g., integer programming, greedy approach) but does not include any explicit pseudocode blocks or sections labeled 'Pseudocode' or 'Algorithm'.
Open Source Code No The paper states that 'The tools developed from this research are currently in daily use at NYC Bike Share LLC' and 'Solutions to the mid-rush rebalancing problem are already in use in New York and the tools developed to solve the Overnight Rebalancing Problem are currently being integrated into the truck dispatching system.', but it does not provide any links to open-source code or explicitly state that the code will be made publicly available.
Open Datasets No We tested the integer programming approaches on real world instances gathered from actual system data. We took a series of system snapshots of the system state at the 8pm start of the overnight shift during June 2013. There is no mention of this data being publicly available or a link to it.
Dataset Splits No The paper discusses using 'real world instances' for experiments but does not provide specific details on how the data was split into training, validation, or test sets (e.g., percentages, sample counts, or explicit splitting methodology).
Hardware Specification Yes The experiments were run on Linux machines with 2 Intel x5690s running at 3.46GHZ, a 128GB SSD and 96GB RAM.
Software Dependencies No The paper states 'We implemented the IP in Gurobi (Gurobi Optimization 2014)'. While 'Gurobi' is a specific solver, '2014' refers to the publication year of its reference manual, not a specific version number for the software (e.g., 6.0 or 6.5).
Experiment Setup Yes We ran the IP with a 900 second cutoff and restricted the greedy to use only 300 seconds for each greedy call to the IP. We implemented the above model in Gurobi, by using a bisection search over the space of possible objective functions we were able to solve instances from New York in under a minute.