linfa-datasets 0.5.1

//! # Datasets
//!
//! `linfa-datasets` provides a collection of commonly used datasets ready to be used in tests and examples.
//!
//! ## The Big Picture
//!
//! `linfa-datasets` is a crate in the [`linfa`](https://crates.io/crates/linfa) ecosystem, an effort to create a toolkit for classical Machine Learning implemented in pure Rust, akin to Python's `scikit-learn`.
//!
//! ## Current State
//!
//! Currently the following datasets are provided:
//!
//! | Name | Description | #samples, #features, #targets | Targets | Reference |
//! | :--- | :--- | :---| :--- | :--- |
//! | iris | The Iris dataset provides samples of flower properties, belonging to three different classes. Only two of them are linearly separable. It was introduced by Ronald Fisher in 1936 as an example for linear discriminant analysis. |  150, 4, 1 | Multi-class classification | [here](https://archive.ics.uci.edu/ml/datasets/iris) |
//! | winequality | The winequality dataset measures different properties of wine, such as acidity, and gives a scoring from 3 to 8 in quality. It was collected in the north of Portugal. | 441, 10, 1 | Multi-class classification | [here](https://archive.ics.uci.edu/ml/datasets/wine+quality) |
//! | diabetes | The diabetes dataset gives samples of human biological measures, such as BMI, age, blood measures, and tries to predict the progression of diabetes. | 1599, 11, 1 | Regression | [here](https://www4.stat.ncsu.edu/~boos/var.select/diabetes.html) |
//! | linnerud | The linnerud dataset contains samples from 20 middle-aged men in a fitness club. Their physical capability, as well as biological measures are related. | 20, 3, 3 | Regression | [here](https://core.ac.uk/download/pdf/20641325.pdf) |
//!
//! The purpose of this crate is to faciliate dataset loading and make it as simple as possible. Loaded datasets are returned as a
//! [linfa::Dataset] structure with named features.
//!
//! ## Using a dataset
//!
//! To use one of the provided datasets in your project add the `linfa-datasets` crate to your `Cargo.toml` and enable the corresponding feature:
//! ```ignore
//! linfa-datasets = { version = "0.3.1", features = ["winequality"] }
//! ```
//!
//! You can then use the dataset in your working code:
//! ```rust
//! let (train, valid) = linfa_datasets::winequality()
//!     .split_with_ratio(0.8);
//! ```

use csv::ReaderBuilder;
use flate2::read::GzDecoder;
use linfa::Dataset;
use ndarray::prelude::*;
use ndarray_csv::Array2Reader;

#[cfg(any(
    feature = "iris",
    feature = "diabetes",
    feature = "winequality",
    feature = "linnerud"
))]
fn array_from_buf(buf: &[u8]) -> Array2<f64> {
    // unzip file
    let file = GzDecoder::new(buf);
    // create a CSV reader with headers and `,` as delimiter
    let mut reader = ReaderBuilder::new()
        .has_headers(true)
        .delimiter(b',')
        .from_reader(file);

    // extract ndarray
    reader.deserialize_array2_dynamic().unwrap()
}

#[cfg(feature = "iris")]
/// Read in the iris-flower dataset from dataset path.
// The `.csv` data is two dimensional: Axis(0) denotes y-axis (rows), Axis(1) denotes x-axis (columns)
pub fn iris() -> Dataset<f64, usize> {
    let data = include_bytes!("../data/iris.csv.gz");
    let array = array_from_buf(&data[..]);

    let (data, targets) = (
        array.slice(s![.., 0..4]).to_owned(),
        array.column(4).to_owned(),
    );

    let feature_names = vec!["sepal length", "sepal width", "petal length", "petal width"];

    Dataset::new(data, targets)
        .map_targets(|x| *x as usize)
        .with_feature_names(feature_names)
}

#[cfg(feature = "diabetes")]
/// Read in the diabetes dataset from dataset path
pub fn diabetes() -> Dataset<f64, f64> {
    let data = include_bytes!("../data/diabetes_data.csv.gz");
    let data = array_from_buf(&data[..]);

    let targets = include_bytes!("../data/diabetes_target.csv.gz");
    let targets = array_from_buf(&targets[..]).column(0).to_owned();

    let feature_names = vec![
        "age",
        "sex",
        "body mass index",
        "blood pressure",
        "t-cells",
        "low-density lipoproteins",
        "high-density lipoproteins",
        "thyroid stimulating hormone",
        "lamotrigine",
        "blood sugar level",
    ];

    Dataset::new(data, targets).with_feature_names(feature_names)
}

#[cfg(feature = "winequality")]
/// Read in the winequality dataset from dataset path
pub fn winequality() -> Dataset<f64, usize> {
    let data = include_bytes!("../data/winequality-red.csv.gz");
    let array = array_from_buf(&data[..]);

    let (data, targets) = (
        array.slice(s![.., 0..11]).to_owned(),
        array.column(11).to_owned(),
    );

    let feature_names = vec![
        "fixed acidity",
        "volatile acidity",
        "citric acid",
        "residual sugar",
        "chlorides",
        "free sulfur dioxide",
        "total sulfur dioxide",
        "density",
        "pH",
        "sulphates",
        "alcohol",
    ];

    Dataset::new(data, targets)
        .map_targets(|x| *x as usize)
        .with_feature_names(feature_names)
}

#[cfg(feature = "linnerud")]
/// Read in the physical exercise dataset from dataset path.
///
/// Linnerud dataset contains 20 samples collected from 20 middle-aged men in a fitness club.
///
/// ## Features:
/// 3 exercises measurements: Chins, Situps, Jumps
///
/// ## Targets:
/// 3 physiological measurements: Weight, Waist, Pulse
///
/// # Reference:
/// Tenenhaus (1998). La regression PLS: theorie et pratique. Paris: Editions Technip. Table p 15.
pub fn linnerud() -> Dataset<f64, f64> {
    let input_data = include_bytes!("../data/linnerud_exercise.csv.gz");
    let input_array = array_from_buf(&input_data[..]);

    let output_data = include_bytes!("../data/linnerud_physiological.csv.gz");
    let output_array = array_from_buf(&output_data[..]);

    let feature_names = vec!["Chins", "Situps", "Jumps"];

    Dataset::new(input_array, output_array).with_feature_names(feature_names)
}

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

    #[cfg(feature = "iris")]
    #[test]
    fn test_iris() {
        let ds = iris();

        // check that we have the right amount of data
        assert_eq!((ds.nsamples(), ds.nfeatures(), ds.ntargets()), (150, 4, 1));

        // check for feature names
        assert_eq!(
            ds.feature_names(),
            &["sepal length", "sepal width", "petal length", "petal width"]
        );

        // check label frequency
        assert_abs_diff_eq!(
            ds.label_frequencies()
                .into_iter()
                .map(|b| b.1)
                .collect::<Array1<_>>(),
            array![50., 50., 50.]
        );

        // perform correlation analysis and assert that petal length and width are correlated
        let _pcc = ds.pearson_correlation_with_p_value(100);
        // TODO: wait for pearson correlation to accept rng
        // assert_abs_diff_eq!(pcc.get_p_values().unwrap()[5], 0.04, epsilon = 0.04);

        // get the mean per feature
        let mean_features = ds.records().mean_axis(Axis(0)).unwrap();
        assert_abs_diff_eq!(
            mean_features,
            array![5.84, 3.05, 3.75, 1.20],
            epsilon = 0.01
        );
    }

    #[cfg(feature = "diabetes")]
    #[test]
    fn test_diabetes() {
        let ds = diabetes();

        // check that we have the right amount of data
        assert_eq!((ds.nsamples(), ds.nfeatures(), ds.ntargets()), (441, 10, 1));

        // perform correlation analysis and assert that T-Cells and low-density lipoproteins are
        // correlated
        let _pcc = ds.pearson_correlation_with_p_value(100);
        //assert_abs_diff_eq!(pcc.get_p_values().unwrap()[30], 0.02, epsilon = 0.02);

        // get the mean per feature, the data should be normalized
        let mean_features = ds.records().mean_axis(Axis(0)).unwrap();
        assert_abs_diff_eq!(mean_features, Array1::zeros(10), epsilon = 0.005);
    }

    #[cfg(feature = "winequality")]
    #[test]
    fn test_winequality() {
        use approx::abs_diff_eq;

        let ds = winequality();

        // check that we have the right amount of data
        assert_eq!(
            (ds.nsamples(), ds.nfeatures(), ds.ntargets()),
            (1599, 11, 1)
        );

        // check for feature names
        let feature_names = vec![
            "fixed acidity",
            "volatile acidity",
            "citric acid",
            "residual sugar",
            "chlorides",
            "free sulfur dioxide",
            "total sulfur dioxide",
            "density",
            "pH",
            "sulphates",
            "alcohol",
        ];
        assert_eq!(ds.feature_names(), feature_names);

        // check label frequency
        let compare_to = vec![
            (5, 681.0),
            (7, 199.0),
            (6, 638.0),
            (8, 18.0),
            (3, 10.0),
            (4, 53.0),
        ];

        let freqs = ds.label_frequencies();
        assert!(compare_to.into_iter().all(|(key, val)| {
            freqs
                .get(&key)
                .map(|x| abs_diff_eq!(*x, val))
                .unwrap_or(false)
        }));

        // perform correlation analysis and assert that fixed acidity and citric acid are
        // correlated
        let _pcc = ds.pearson_correlation_with_p_value(100);
        //assert_abs_diff_eq!(pcc.get_p_values().unwrap()[1], 0.05, epsilon = 0.05);
    }

    #[cfg(feature = "linnerud")]
    #[test]
    fn test_linnerud() {
        let ds = linnerud();

        // check that we have the right amount of data
        assert_eq!((ds.nsamples(), ds.nfeatures(), ds.ntargets()), (20, 3, 3));

        // check for feature names
        let feature_names = vec!["Chins", "Situps", "Jumps"];
        assert_eq!(ds.feature_names(), feature_names);

        // get the mean per target: Weight, Waist, Pulse
        let mean_targets = ds.targets().mean_axis(Axis(0)).unwrap();
        assert_abs_diff_eq!(mean_targets, array![178.6, 35.4, 56.1]);
    }
}