datarust 0.6.5

Scikit-learn-style preprocessing and classical ML in Rust
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132

//! Enum wrapper for target (supervised) categorical transformers, enabling
//! serialization of [`ColumnTransformer`] under the `serde` feature.
//!
//! [`ColumnTransformer`]: crate::compose::ColumnTransformer

use crate::encoder::TargetEncoder;
use crate::error::Result;
use crate::matrix::{Matrix, StrMatrix};
use crate::traits::{FeatureNames, TargetTransformer};

/// Concrete target transformer variant, wrapping a [`TargetEncoder`] for use
/// in [`crate::ColumnTransformer`].
///
/// Each variant wraps a single supervised encoder that implements
/// [`TargetTransformer`] (i.e. operates on `(StrMatrix, &[f64]) -> Matrix`).
#[derive(Debug, Clone)]
#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
pub enum TargetTransformerKind {
    /// Wraps a `TargetEncoder`.
    TargetEncoder(TargetEncoder),
}

impl TargetTransformer for TargetTransformerKind {
    fn name(&self) -> &'static str {
        match self {
            Self::TargetEncoder(t) => t.name(),
        }
    }

    fn fit(&mut self, x: &StrMatrix, y: &[f64]) -> Result<()> {
        match self {
            Self::TargetEncoder(t) => t.fit(x, y),
        }
    }

    fn transform(&self, x: &StrMatrix) -> Result<Matrix> {
        match self {
            Self::TargetEncoder(t) => t.transform(x),
        }
    }

    fn is_fitted(&self) -> bool {
        match self {
            Self::TargetEncoder(t) => t.is_fitted(),
        }
    }

    fn inverse_transform(&self, y: &Matrix) -> Result<StrMatrix> {
        match self {
            Self::TargetEncoder(t) => t.inverse_transform(y),
        }
    }
}

impl FeatureNames for TargetTransformerKind {
    fn feature_names_out(&self, input_features: Option<&[String]>) -> Vec<String> {
        match self {
            Self::TargetEncoder(t) => t.feature_names_out(input_features),
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    fn sample_str() -> StrMatrix {
        StrMatrix::from_column(["A", "A", "B", "B"]).unwrap()
    }

    #[test]
    fn name_delegates_to_inner() {
        let kind = TargetTransformerKind::TargetEncoder(TargetEncoder::new(0.0).unwrap());
        assert_eq!(kind.name(), TargetEncoder::new(0.0).unwrap().name());
    }

    #[test]
    fn fit_transform_dispatches_to_target_encoder() {
        let s = sample_str();
        // Targets: A→0.0, B→1.0. With smoothing=0 the encoding is the category mean.
        let y = vec![0.0, 0.0, 1.0, 1.0];
        let mut kind = TargetTransformerKind::TargetEncoder(TargetEncoder::new(0.0).unwrap());
        assert!(!kind.is_fitted());
        let out = kind.fit_transform(&s, &y).unwrap();
        assert_eq!(out.nrows(), 4);
        assert_eq!(out.ncols(), 1);
        // A maps to mean 0.0, B to mean 1.0.
        assert!((out.get(0, 0) - 0.0).abs() < 1e-9);
        assert!((out.get(2, 0) - 1.0).abs() < 1e-9);
        assert!(kind.is_fitted());
    }

    #[test]
    fn transform_before_fit_errors() {
        let s = sample_str();
        let kind = TargetTransformerKind::TargetEncoder(TargetEncoder::new(0.0).unwrap());
        let err = kind.transform(&s).unwrap_err();
        assert!(matches!(err, crate::error::DatarustError::NotFitted(_)));
    }

    #[test]
    fn smoothing_pulls_toward_global_mean() {
        // With heavy smoothing, every category encoding approaches the global mean.
        let s = sample_str();
        let y = vec![0.0, 0.0, 1.0, 1.0];
        let mut kind = TargetTransformerKind::TargetEncoder(TargetEncoder::new(1000.0).unwrap());
        let out = kind.fit_transform(&s, &y).unwrap();
        // Global mean = 0.5; with smoothing=1000 both categories are pulled close to it.
        let global = 0.5;
        for i in 0..s.nrows() {
            assert!(
                (out.get(i, 0) - global).abs() < 1e-3,
                "row {} got {} expected ~{}",
                i,
                out.get(i, 0),
                global
            );
        }
    }

    #[test]
    fn feature_names_delegates_to_inner() {
        // The wrapper delegates to the inner encoder; mappings are populated
        // only after fit, so we fit first.
        let s = sample_str();
        let y = vec![0.0, 0.0, 1.0, 1.0];
        let mut kind = TargetTransformerKind::TargetEncoder(TargetEncoder::new(0.0).unwrap());
        kind.fit(&s, &y).unwrap();
        let names = kind.feature_names_out(Some(&["cat".into()]));
        assert_eq!(names, vec!["cat".to_string()]);
    }
}