evaluate.py

import warnings
warnings.filterwarnings("ignore")

import pandas as pd
import numpy as np

import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.model_selection import train_test_split
from scipy.stats import pearsonr, spearmanr

from env import get_db_url
import wrangle as w
import prepare as p
import matplotlib as mpl
mpl.rcParams["axes.formatter.useoffset"] = False
#import sklearn mean_square_error
from sklearn.metrics import mean_squared_error
#import r2_score from sklearn
from sklearn.metrics import r2_score
import sklearn.preprocessing


def plot_residuals(y, yhat):
    residuals = y - yhat
    
    plt.scatter(x=y, y=residuals)
    plt.xlabel('Home Value')
    plt.ylabel('Residuals')
    plt.title('Residual vs Home Value Plot')
    plt.show()


def regression_errors(y, yhat):
    MSE = mean_squared_error(y, yhat)
    SSE = MSE * len(y)
    RMSE = MSE**.5
    
    ESS = ((yhat - y.mean())**2).sum()
    TSS = ESS + SSE
    
    return SSE, ESS, TSS, MSE, RMSE


def baseline_mean_errors(y):
    baseline = np.repeat(y.mean(), len(y))
    
    MSE = mean_squared_error(y, baseline)
    SSE = MSE * len(y)
    RMSE = MSE**.5
    
    return SSE, MSE, RMSE

def better_than_baseline(y, yhat):
    SSE, ESS, TSS, MSE, RMSE = regression_errors(y, yhat)
    
    SSE_baseline, MSE_baseline, RMSE_baseline = baseline_mean_errors(y)
    
    if SSE < SSE_baseline:
        print('My OSL model performs better than baseline')
    else:
        print('My OSL model performs worse than baseline. :( )')