Compute rankings in Python.
Project description
Ranky
Compute rankings in Python.
Get started
pip install ranky
import ranky as rk
Read the documentation.
Main functions
The main functionalities include scoring metrics (e.g. accuracy, roc auc), rank metrics (e.g. Kendall Tau, Spearman correlation), ranking systems (e.g. Majority judgement, Kemeny-Young method) and some measurements (e.g. Kendall's W coefficient of concordance).
Most functions takes as input 2-dimensional numpy.array
or pandas.DataFrame
objects. DataFrame are the best to keep track of the names of each data point.
Let's consider the following preference matrix:
Each row is a candidate and each column is a judge. Here is the results of rk.borda(matrix)
, computing the mean rank of each candidate:
We can see that candidate2 has the best average rank among the four judges.
Let's display it using rk.show(rk.borda(matrix))
:
Ranking systems
The rank aggregation methods available include:
- Random Dictator:
rk.dictator(m)
- Score Voting (mean):
rk.score(m)
- Borda Count (average rank):
rk.borda(m)
- Majority Judgement (median):
rk.majority(m)
- Pairwise methods. Copeland's method:
rk.pairwise(m)
, Success rate:rk.pairwise(m, wins=rk.success_rate)
and more. You can specify your own "wins" function or select one from therk.duel
module. - Optimal rank aggregation using any rank metric:
rk.center(m)
,rk.center(m, method='kendalltau')
. Solver used [1]. - (Kemeny-Young method is optimal rank aggregation using Kendall's tau as metric.)
- (Optimal rank aggregation using Spearman correlation as metric is equivalent to Borda count.)
Metrics
You can use any_metric(a, b, method)
to call a metric from any of the three categories below.
-
Scoring metrics:
rk.metric(y_true, y_pred, method='accuracy')
. Methods include:['accuracy', 'balanced_accuracy', 'precision', 'average_precision', 'brier', 'f1_score', 'mxe', 'recall', 'jaccard', 'roc_auc', 'mse', 'rmse', 'sar']
-
Rank correlation coefficients:
rk.corr(r1, r2, method='spearman')
. Methods include:['kendalltau', 'spearman', 'pearson']
-
Rank distances:
rk.dist(r1, r2, method='levenshtein')
. Methods include:['hamming', 'levenshtein', 'kendall', 'winner', 'euclidean']
To add: general edit distances, kemeny distance, regression metrics...
Visualizations
- Use
rk.show
to visualize preference matrix (2D) or ranking ballots (1D).
>>> rk.show(m)
>>> rk.show(m['judge1'])
- Use
rk.mds
, to visualize (in 2D or 3D) the points in a given metric space. Seerk.scatterplot
documentation for display arguments.
>>> rk.mds(m, method='euclidean')
>>> rk.mds(m, method='spearman', axis=1)
-
You can use
rk.tsne
similarly tork.mds
. -
Use
rk.critical_difference
to plot a critical difference diagram, comparing candidates' performance and grouping them by statistical equivalence. Such diagrams can be seen in [2, 3].
>>> rk.critical_difference(m, comparison_func=rk.bayes_wins)
- Show Condorcet graphs using
rk.show_graph(graph)
, based on [4].
Other
- Rank,
rk.rank
, convert a 1D score ballot into a ranking. - Bootstrap,
rk.bootstrap
, sample a given axis. - Consensus,
rk.consensus
, check if ranking exactly agree. - Concordance, ,
rk.concordance
, mean rank distance between all judges of a preference matrix. - Centrality,
rk.centrality
, mean rank distance between a ranking and a preference matrix. - Kendall's W,
rk.kendall_w
, coefficient of concordance. - Utility:
read_codalab_csv
to parse a CSV generated by Codalab representing a leaderboard into apandas.DataFrame
.
References
Please cite ranky in your publications if this is useful for your research. Here is an example BibTeX entry:
@misc{pavao2020ranky,
title={ranky},
author={Adrien Pavao},
year={2020},
howpublished={\url{https://github.com/didayolo/ranky}},
}
[1] Storn R. and Price K., Differential Evolution - a Simple and Efficient Heuristic for Global Optimization over Continuous Spaces, Journal of Global Optimization, 1997, 11, 341 - 359.
[2] Janez Demsar, Statistical Comparisons of Classifiers over Multiple Data Sets, 7(Jan):1--30, 2006.
[3] H. Ismail Fawaz, G. Forestier, J. Weber, L. Idoumghar, P. Muller, Deep learning for time series classification: a review, Data Mining and Knowledge Discovery, 2018.
[4] Aric A. Hagberg, Daniel A. Schult and Pieter J. Swart, “Exploring network structure, dynamics, and function using NetworkX”, in Proceedings of the 7th Python in Science Conference (SciPy2008), Gäel Varoquaux, Travis Vaught, and Jarrod Millman (Eds), (Pasadena, CA USA), pp. 11–15, Aug 2008.
License
Copyright (c) 2020-2021, Adrien PAVAO. This software is released under the Apache License 2.0 (the "License"); you may not use the software except in compliance with the License.
The text of the Apache License 2.0 can be found online at: http://www.opensource.org/licenses/apache2.0.php
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
Built Distribution
File details
Details for the file ranky-0.3.5-py3-none-any.whl
.
File metadata
- Download URL: ranky-0.3.5-py3-none-any.whl
- Upload date:
- Size: 26.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.12.4
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 01220338dad9902d00e2339a15b9922113cfeadc4fa833510556a148fe6aa06f |
|
MD5 | 23f801df0b43c131a99ee0ba7a052e1c |
|
BLAKE2b-256 | 779edc2daa96ce8a92a04381e88854d65062eb2fe270c32d970dd58136323aeb |