Python for Data Science

Chapter 3 - Regression Models

Segment 2 - Multiple linear regression

import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
from pylab import rcParams

import sklearn
from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import scale

%matplotlib inline
rcParams['figure.figsize'] = 5, 4

import seaborn as sb
sb.set_style('whitegrid')
from collections import Counter

(Multiple) linear regression on the enrollment data

address = '~/Data/enrollment_forecast.csv'

enroll = pd.read_csv(address)
enroll.columns = ['year', 'roll', 'unem', 'hgrad', 'inc']
enroll.head()

	year	roll	unem	hgrad	inc
0	1	5501	8.1	9552	1923
1	2	5945	7.0	9680	1961
2	3	6629	7.3	9731	1979
3	4	7556	7.5	11666	2030
4	5	8716	7.0	14675	2112

sb.pairplot(enroll)

<seaborn.axisgrid.PairGrid at 0x7f9c02deb7f0>

print(enroll.corr())

           year      roll      unem     hgrad       inc
year   1.000000  0.900934  0.378305  0.670300  0.944287
roll   0.900934  1.000000  0.391344  0.890294  0.949876
unem   0.378305  0.391344  1.000000  0.177376  0.282310
hgrad  0.670300  0.890294  0.177376  1.000000  0.820089
inc    0.944287  0.949876  0.282310  0.820089  1.000000

enroll_data = enroll[['unem', 'hgrad']].values

enroll_target = enroll[['roll']].values

enroll_data_names = ['unem', 'hgrad']

X, y = scale(enroll_data), enroll_target

Checking for missing values

missing_values = X==np.NAN
X[missing_values == True]

array([], dtype=float64)

LinReg = LinearRegression(normalize=True)

LinReg.fit(X, y)

print(LinReg.score(X, y))

0.8488812666133723