Boston_linear.py

# Boston_linear.py
# Fit a linear model to the Boston Housing data and display the predictor importance

# modules
import numpy as np
from tabulate import tabulate
from sklearn.datasets import load_boston
from sklearn.metrics import mean_squared_error, r2_score
from sklearn.linear_model import LinearRegression


def main():
    # load dataset
    data = load_boston()
    dataX = data.data
    dataY = data.target
    feature_names = data.feature_names

    # fit the linear model
    model_linear = LinearRegression()
    model_linear.fit(dataX, dataY)

    # output fit result
    pred = model_linear.predict(dataX)
    print("-" * 80)
    print("Fit results:", "\n")
    print("Coefficients:")
    print(tabulate(zip(feature_names, model_linear.coef_), tablefmt="simple"), "\n\n")
    print("The training MSE:", mean_squared_error(dataY, pred))
    print("The coefficients of determination:", r2_score(dataY, pred))
    print("-" * 80, "\n\n")

    # importance of predictors
    coef_abs = np.abs(model_linear.coef_)
    temp = sorted(coef_abs, reverse=True)
    ordered_index = [np.where(coef_abs == i)[0][0] for i in temp]
    ordered_features = feature_names[ordered_index]
    ordered_coef = np.round(model_linear.coef_[ordered_index], 4)
    ordered_coef_abs = np.abs(ordered_coef)

    # output
    print("Feature Importance: from greatest to least", "\n")
    print(
        tabulate(
            np.array([ordered_features, ordered_coef_abs, ordered_coef]).T,
            headers=["Feature Name", "Abs(Coefficients)", "Coefficients"],
            tablefmt="github",
        )
    )


if __name__ == "__main__":
    main()