PrivateJobMatch

A repository containing many components used to build and test PrivateJobMatch using Python 2.7

Thesis: https://arxiv.org/abs/1905.04564

Long-Paper @ RecSys 2019: x

---

Dataset Generator

• Small script to generate synthetic candidate-employer preferences using the function createDatasets in Dataset_Generator.py
• Preferences are reasonably decided based on a utility function (assigned type).
• Specify the number of candidates and employers, n and m respectively.

---

Job Market Simulator

• Code for simulating a job market.
• Employers offer jobs to candidates using a descending order of candidate priority, which can be inputted in the form of:
  • Raw preferences (as used in today's decentralized job market).
  • MMDAA match output (centralized job market).

---

Low-Rank Matrix Factorization

• Code for performing LMF on a sparse dataset. See LMF_Runner.py
• Included a utility function, sparsifyDataset, for converting a dense dataset into a sparse dataset.

---

Multi-Match Deferred Acceptance Algorithm

• MMDAA.py contains functions for various tasks.
• setUpInputFile will create a text file in the required format/structure for parsing when setting up the data structures (preprocessing) for running the MMDAA.
  • See exampleInput.txt for required preference structure.
  • Takes two dense preference files (CSV format) as input. Outputs a text file.
• mainRunner loads the input text file that holds the required preference data, and runs the MMDAA. Saves the output in CSV files, as well as a text file.
• runMetric is a function used to select a metric to calculate, given match outputs and preference inputs.

---

Results

• A folder containing experimental results (displacement, withholdings, vacancy) in the form of spreadsheets for various datasets.

---