featrs 0.2.0

Feature engineering library for Rust, inspired by scikit-learn
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
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155

//! Feature binarization.
//!
//! [`Binarizer`] thresholds numeric features: values above the threshold
//! become `1.0`, others become `0.0`.

use crate::traits::{Error, Fit, Result, Transform};
use polars::prelude::*;

/// Binarize data according to a threshold.
///
/// Values `> threshold` become `1.0`; all others become `0.0`.
///
/// # Example
///
/// ```rust
/// use featrs::preprocessing::binarizer::Binarizer;
/// use featrs::traits::{Fit, Transform};
///
/// let mut b = Binarizer::new(0.5);
/// # let df = polars::prelude::DataFrame::new(0usize, vec![]).unwrap();
/// // b.fit(df.clone(), target)?;
/// // let binarized = b.transform(df)?;
/// ```
pub struct Binarizer {
    fitted: bool,
    threshold: f64,
}

impl Binarizer {
    /// Create a new `Binarizer` with the given threshold.
    ///
    /// Values strictly greater than `threshold` are set to `1.0`.
    pub fn new(threshold: f64) -> Self {
        Self {
            fitted: false,
            threshold,
        }
    }
}

impl Default for Binarizer {
    fn default() -> Self {
        Self::new(0.0)
    }
}

impl Fit<DataFrame, DataFrame> for Binarizer {
    type Output = ();

    fn fit(&mut self, x: DataFrame, _y: DataFrame) -> Result<()> {
        if x.width() == 0 {
            return Err(Error::InvalidInput(
                "Binarizer.fit received a DataFrame with 0 columns. \
                 Provide at least one column."
                    .into(),
            ));
        }
        self.fitted = true;
        Ok(())
    }
}

impl Transform<DataFrame> for Binarizer {
    type Output = DataFrame;

    fn transform(&self, x: DataFrame) -> Result<DataFrame> {
        if !self.fitted {
            return Err(Error::NotFitted(
                "Binarizer has not been fitted. \
                 Call .fit(dataframe, target) before .transform()."
                    .into(),
            ));
        }

        let col_names: Vec<String> = x
            .get_column_names()
            .iter()
            .filter_map(|name| {
                let col = x.column(name).ok()?;
                if col.dtype() == &DataType::Float64 {
                    Some(name.to_string())
                } else {
                    None
                }
            })
            .collect();

        let mut out = x.clone();

        for name in &col_names {
            let s = out.column(name.as_str()).unwrap();
            let ca = s.f64().unwrap();
            let binarized: ChunkedArray<Float64Type> = ca
                .iter()
                .map(|opt| opt.map(|v| if v > self.threshold { 1.0 } else { 0.0 }))
                .collect();
            out.replace(name.as_str(), binarized.into_series().into())
                .unwrap();
        }

        Ok(out)
    }
}

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

    #[test]
    fn test_binarizer_default() {
        let mut b = Binarizer::default();
        let a = Column::from(Series::new("x".into(), &[-1.0f64, 0.0, 2.0]));
        let df = DataFrame::new(3, vec![a]).unwrap();
        let y = df.clone();

        b.fit(df.clone(), y).unwrap();
        let result = b.transform(df).unwrap();

        let vals: Vec<f64> = result
            .column("x")
            .unwrap()
            .f64()
            .unwrap()
            .iter()
            .flatten()
            .collect();
        assert_relative_eq!(vals[0], 0.0, epsilon = 1e-6);
        assert_relative_eq!(vals[1], 0.0, epsilon = 1e-6);
        assert_relative_eq!(vals[2], 1.0, epsilon = 1e-6);
    }

    #[test]
    fn test_binarizer_custom_threshold() {
        let mut b = Binarizer::new(5.0);
        let a = Column::from(Series::new("x".into(), &[1.0f64, 5.0, 10.0]));
        let df = DataFrame::new(3, vec![a]).unwrap();
        let y = df.clone();

        b.fit(df.clone(), y).unwrap();
        let result = b.transform(df).unwrap();

        let vals: Vec<f64> = result
            .column("x")
            .unwrap()
            .f64()
            .unwrap()
            .iter()
            .flatten()
            .collect();
        assert_relative_eq!(vals[0], 0.0, epsilon = 1e-6);
        assert_relative_eq!(vals[1], 0.0, epsilon = 1e-6);
        assert_relative_eq!(vals[2], 1.0, epsilon = 1e-6);
    }
}