Module linear_regression 

Linear Regression

Linear regression is a very straightforward approach for predicting a quantitative response \(y\) on the basis of a linear combination of explanatory variables \(X\). Linear regression assumes that there is approximately a linear relationship between \(X\) and \(y\). Formally, we can write this linear relationship as

\[y \approx \beta_0 + \sum_{i=1}^n \beta_iX_i + \epsilon\]

where \(\epsilon\) is a mean-zero random error term and the regression coefficients \(\beta_0, \beta_0, … \beta_n\) are unknown, and must be estimated.

While regression coefficients can be estimated directly by solving

\[\hat{\beta} = (X^TX)^{-1}X^Ty \]

the \((X^TX)^{-1}\) term is both computationally expensive and numerically unstable. An alternative approach is to use a matrix decomposition to avoid this operation. smartcore uses SVD and QR matrix decomposition to find estimates of \(\hat{\beta}\). The QR decomposition is more computationally efficient and more numerically stable than calculating the normal equation directly, but does not work for all data matrices. Unlike the QR decomposition, all matrices have an SVD decomposition.

Example:

use smartcore::linalg::basic::matrix::DenseMatrix;
use smartcore::linear::linear_regression::*;

// Longley dataset (https://www.statsmodels.org/stable/datasets/generated/longley.html)
let x = DenseMatrix::from_2d_array(&[
              &[234.289, 235.6, 159.0, 107.608, 1947., 60.323],
              &[259.426, 232.5, 145.6, 108.632, 1948., 61.122],
              &[258.054, 368.2, 161.6, 109.773, 1949., 60.171],
              &[284.599, 335.1, 165.0, 110.929, 1950., 61.187],
              &[328.975, 209.9, 309.9, 112.075, 1951., 63.221],
              &[346.999, 193.2, 359.4, 113.270, 1952., 63.639],
              &[365.385, 187.0, 354.7, 115.094, 1953., 64.989],
              &[363.112, 357.8, 335.0, 116.219, 1954., 63.761],
              &[397.469, 290.4, 304.8, 117.388, 1955., 66.019],
              &[419.180, 282.2, 285.7, 118.734, 1956., 67.857],
              &[442.769, 293.6, 279.8, 120.445, 1957., 68.169],
              &[444.546, 468.1, 263.7, 121.950, 1958., 66.513],
              &[482.704, 381.3, 255.2, 123.366, 1959., 68.655],
              &[502.601, 393.1, 251.4, 125.368, 1960., 69.564],
              &[518.173, 480.6, 257.2, 127.852, 1961., 69.331],
              &[554.894, 400.7, 282.7, 130.081, 1962., 70.551],
         ]).unwrap();

let y: Vec<f64> = vec![83.0, 88.5, 88.2, 89.5, 96.2, 98.1, 99.0,
          100.0, 101.2, 104.6, 108.4, 110.8, 112.6, 114.2, 115.7, 116.9];

let lr = LinearRegression::fit(&x, &y,
            LinearRegressionParameters::default().
            with_solver(LinearRegressionSolverName::QR)).unwrap();

let y_hat = lr.predict(&x).unwrap();

References:

• “Pattern Recognition and Machine Learning”, C.M. Bishop, Linear Models for Regression
• “An Introduction to Statistical Learning”, James G., Witten D., Hastie T., Tibshirani R., 3. Linear Regression
• “Numerical Recipes: The Art of Scientific Computing”, Press W.H., Teukolsky S.A., Vetterling W.T, Flannery B.P, 3rd ed., Section 15.4 General Linear Least Squares

Structs

LinearRegression

Linear Regression

LinearRegressionParameters

Linear Regression parameters

LinearRegressionSearchParameters

Linear Regression grid search parameters

LinearRegressionSearchParametersIterator

Linear Regression grid search iterator

Enums

LinearRegressionSolverName

Approach to use for estimation of regression coefficients. QR is more efficient but SVD is more stable.