pabutools 0.8

Implementation of PB voting rules and tools for reading Pabulib datasets

pabutools---tools for participatory budgeting

Pabutools provides two submodules: model and rules.

model

Provides the classes representing voters, candidates and election instances.

• Voter(id, sex=None, age=None, subunits=set()) with the following attributes:

  • id: unique ID of the voter,
  • sex: 'M' (male), 'F' (female) or None (no information),
  • age: a natural number or None (no information),
  • subunits: a set of strings (subunits which the voter belongs to).

• Candidate(id, cost, name=None, subunit=None) with the following attributes:

  • id: unique ID of the candidate,
  • cost: a natural number,
  • name: string or None (no information),
  • subunit: string or None (no information or a citywide project).

• Election(name=None, voters=None, profile=None, budget=0, subunits = None) with the following attributes:

  • voters: a set of Voter instances,
  • profile: a dict where keys are Candidate instances and values are dicts in which keys are Voter instances and values are natural numbers (voters' utilities over candidates; if a voter has utility 0, she can be skipped),
  • budget: a natural number,
  • subunits: a set of strings (subunits within the instance).

  and the following methods:

  • read_from_file(pattern): takes as an input the pattern of the filepaths of Pabulib files and fills the data of the newly created Election instance with them (e.g. calling Election().read_from_files('pabulib/poland_warszawa_2021*') will create a PB election out of all Pabulib files under the provided path starting with 'poland_warszawa_2021'). If the pattern fits more than one file, they should all be from the same country, unit and year. By default, approval utilities are considered to be binary.
  • cost_to_binary_utilities(): if the utilities are cost utilities, they are converted to binary
  • binary_to_cost_utilities(): if the utilities are binary, they are converted to cost utilities

rules

Provides the implementation of a number of voting rules for PB. All the methods take as an argument an election instance and return a set of elected candidates.

• utilitarian_greedy(e : Election): the simple greedy algorithm,
• phragmen(e : Election): the Phragmen's Sequential Method,
• equal_shares(e : Election, completion=None): the method of equal shares. The 'completion' parameter can take the following values:
  • None: no completion (default option).
  • 'binsearch': initial endowments of the voters are set to maximize the exhaustiveness of the elected committee with the use of binary search.
  • 'add1': initial endowments of the voters are increased similarly as above so that the committee is as exhaustive as possible; instead of using binary search, we keep increasing voters' endowments by 1 unit until the outcome is exhaustive or we exceed the original budget. - 'add1_utilitarian': similarly as in add1 completion, but if the outcome is still not exhaustive, it is completed by utilitarian greedy.
  • 'utilitarian_greedy': the completion with utilitarian_greedy.
  • 'phragmen': the completion with the Phragmen's Sequential Method.