https://github.com/sile/fanova

A Rust implementation of fANOVA (functional analysis of variance)
https://github.com/sile/fanova

fanova feature-importance hyperparameter-importance

Last synced: about 1 year ago
JSON representation

A Rust implementation of fANOVA (functional analysis of variance)

• Host: GitHub
• URL: https://github.com/sile/fanova
• Owner: sile
• License: mit
• Created: 2020-05-17T16:25:22.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2022-07-01T10:09:47.000Z (almost 4 years ago)
• Last Synced: 2025-03-25T07:12:44.767Z (over 1 year ago)
• Topics: fanova, feature-importance, hyperparameter-importance
• Language: Rust
• Homepage:
• Size: 140 KB
• Stars: 15
• Watchers: 2
• Forks: 1
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
fanova

=======

[![fanova](https://img.shields.io/crates/v/fanova.svg)](https://crates.io/crates/fanova)

[![Documentation](https://docs.rs/fanova/badge.svg)](https://docs.rs/fanova)

[![Actions Status](https://github.com/sile/fanova/workflows/CI/badge.svg)](https://github.com/sile/fanova/actions)

[![Coverage Status](https://coveralls.io/repos/github/sile/fanova/badge.svg?branch=master)](https://coveralls.io/github/sile/fanova?branch=master)

[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)

A Rust [fANOVA] (functional analysis of variance) implementation.

fANOVA provides a way to calculate feature importance.

Examples

--------

```rust

use fanova::{FanovaOptions, RandomForestOptions};

use rand::{Rng, SeedableRng};

let mut feature1 = Vec::new();

let mut feature2 = Vec::new();

let mut feature3 = Vec::new();

let mut target = Vec::new();

let mut rng = rand::rngs::StdRng::seed_from_u64(0);

for _ in 0..100 {

    let f1 = rng.gen();

    let f2 = rng.gen();

    let f3 = rng.gen();

    let t = f1 + f2 * 2.0 + f3 * 3.0;

    feature1.push(f1);

    feature2.push(f2);

    feature3.push(f3);

    target.push(t);

}

let mut fanova = FanovaOptions::new()

    .random_forest(RandomForestOptions::new().seed(0))

    .fit(vec![&feature1, &feature2, &feature3], &target).unwrap();

let importances = (0..3)

    .map(|i| fanova.quantify_importance(&[i]).mean)

    .collect::>();

assert_eq!(

    importances,

    vec![0.02744461966313835, 0.22991883769286145, 0.6288784011550144]

);

```

References

----------

- [An Efficient Approach for Assessing Hyperparameter Importance][fANOVA]

[fANOVA]: http://proceedings.mlr.press/v32/hutter14.html