model_evaluator.py

# Imports
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

from sklearn.datasets import load_boston
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import cross_val_score

from itertools import combinations
import time

# Load the Boston housing price data
boston = load_boston()
X = pd.DataFrame(boston.data, columns = boston.feature_names)
y = boston['target']
features = list(boston.feature_names)

# Function to generate and evaluate linear regression models from a set of features
def model_evaluator(X, y, features, min_n, max_n):

    # Instantiate the linear regression model
    lm = LinearRegression()
    
    # Set up the model list
    model_library = []
    
    # Loop through all combinations starting with min_n number of features and 
    # ending with max_n number of featues in X
    for i in range(min_n, max_n + 1):
        
        # Get the i combinations from features
        combos = list(combinations(features, i))
        
        # For each combo create a model, generate predictions, score it and save in a dictionary
        for combo in combos:
        
            # Get this list of features
            feat_sub = list(combo)
            
            # Create a new X with a subset of features
            X_sub = X[feat_sub]
            
            # fit the model and generate predictions
            model = lm.fit(X_sub, y)
            pred = model.predict(X_sub)
            
            # score the model
            score_r2 = model.score(X_sub, y) 
            score_cv = cross_val_score(lm, X_sub, y, cv = 5).mean()
            
            # save this model in a list of dictionaries 
            model_library.append({
                'model': 'linear_regr',
                'features': list(combo),
                'CV_R2': np.round(score_cv, 3),
                'R2': np.round(score_r2, 3),
                'intercept': np.round(model.intercept_, 2),
                'coeff' : np.round(model.coef_, 2),
                'no_features' : len(combo),
            })
            
    # return a data frame
    return pd.DataFrame(model_library)