Skip to content

LEYADEV/homeless

 
 

Repository files navigation

"homeless" repo

This repository presents anonymized dataset and experiment results described in paper "Rehabilitating Homeless: Dataset and Key Insights" (arxiv link will follow).

The dataset is based on more than 9 years' experience and data collection of "Nochlezhka" non-government organization helping homeless people.

If you intend to use this dataset for whatever purposes, visit "Nochlezhka" at https://homeless.ru/en/how_to_help/ and donate for the noble cause of helping the homeless.

This repo contains the following files:

  • Appendix.pdf - appendix to the original "Rehabilitating Homeless: Dataset and Key Insights" paper containing supplementary tables with detailed description of dataset features and NaN statistics featurewise.
  • all_events_anonymized.csv - events (visits to humanitarian projects or entering social contracts) for each person (anonymized and represented by unique IDs) with dates.
  • anonymized_contracts.csv - social contracts for each person (anonymized and represented by unique IDs) in more detail - with contract statuses, start dates and end dates.
  • anonymized_ids_map.csv - representation of each person (each row is a person) by a pair of IDs (one for social contracts, one for humanitarian projects - one person may have either one or both IDs).
  • contract_status.csv - description of numerical contract status IDs used in the dataset.
  • contract_types.csv - description of numerical contract type IDs used in the dataset.
  • final_with_onehot_types_all_221120.csv - the dataset "body" - social contract events with all features extracted from "Nochlezhka" database, along with contract statuses and results.
  • nochlezhka_rel_2.ipynb - experiments described in the paper, for the sake of reproducibility.

About

Dataset and experiment results described in paper "Rehabilitating Homeless: Dataset and Key Insights"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 100.0%