In [1]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
In [2]:
wine_df = pd.read_csv('winequality-red.csv')
In [3]:
myX = np.array(wine_df.drop(columns = ['quality']))
In [4]:
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA
In [5]:
sc = StandardScaler()
pca = PCA(n_components=5)
In [6]:
Xstd = sc.fit_transform(myX)
In [7]:
pca.fit(Xstd)
Out[7]:
PCA(n_components=5)
In [8]:
xrange = np.linspace(1, 10, num=10)
In [10]:
plt.plot(xrange[0:5], np.cumsum(pca.explained_variance_ratio_))
plt.xlabel('number of components')
plt.ylabel('cumulative explained variance')
Out[10]:
Text(0, 0.5, 'cumulative explained variance')
In [55]:
plt.plot(xrange[0:5],pca.explained_variance_ratio_)
plt.xlabel('component')
plt.ylabel('explained variance ratio')
plt.show()
In [9]:
pcadf = pd.DataFrame(pca.fit_transform(Xstd))
In [10]:
pca.explained_variance_ratio_
Out[10]:
array([0.28173931, 0.1750827 , 0.1409585 , 0.11029387, 0.08720837])
In [41]:
plt.scatter(pcadf.loc[:,1], pcadf.loc[:,2], c = wine_df['quality'], cmap = 'viridis')
plt.xlabel('PC 2')
plt.ylabel('PC 3')
plt.title('Observations plotted on PCs')
#plt.colorbar()
plt.show()
In [17]:
hilowines = wine_df[(wine_df['quality']>6) | (wine_df['quality']<5)]
In [18]:
corr = wine_df.corr()
In [19]:
import seaborn as sns
In [11]:
comps = pd.DataFrame(pca.components_)
In [12]:
comps.columns = wine_df.columns[0:11]
In [13]:
comps.T.sort_values(by=abs(comps.T.columns[2]), ascending = False).drop(columns = [0,1])
Out[13]:
2 3 4
alcohol 0.471673 -0.122181 0.350681
free sulfur dioxide 0.428793 -0.043538 -0.159152
total sulfur dioxide 0.322415 -0.034577 -0.222465
sulphates 0.279786 0.550872 0.225962
citric acid 0.238247 -0.079418 -0.058573
residual sugar 0.101283 -0.372793 0.732144
pH 0.057697 -0.003788 0.267530
chlorides -0.092614 0.666195 0.246501
fixed acidity -0.123302 -0.229617 -0.082614
density -0.338871 -0.174500 0.157077
volatile acidity -0.449963 0.078960 0.218735
In [1]:
# comp_vars = pd.DataFrame()
# comp_vals = pd.DataFrame()
# for col in comps.T.columns:
#     sortname = 'comp'+str(col)
#     sorted_comp = comps.T.sort_values(by=abs(comps.columns[col]), ascending=False).reset_index().rename(columns = {'index':sortname})
#     comp_vars[sortname] = sorted_comp.loc[:, sortname]
#     comp_vals[sortname] = comps.loc[col]
#     #print(sorted_comp.head())
In [15]:
comp_vars_names = pd.DataFrame(columns = comp_vars.columns)
#counter = 0
for col in comp_vars.columns:
    mylist = comp_vars[col]
    mynames = wine_df.columns[mylist]
    col2name = str(col) + 'values'
    #print(mynames)
    for ii in range(0,3):
        comp_vars_names.loc[ii,col] = mynames[ii]
        #comp_vars_names.loc[ii, col2name] = sorted_comp.loc[comp_vars[]]
    #counter = counter +1
In [129]:
pca.explained_variance_ratio_
Out[129]:
array([0.28173931, 0.1750827 , 0.1409585 , 0.11029387, 0.08720837])
In [17]:
hilowines = wine_df[(wine_df['quality']>6) | (wine_df['quality']<5)]
In [19]:
%matplotlib notebook
In [57]:
fig = plt.figure()
ax = fig.add_subplot(projection='3d')
ax.scatter(pcadf.loc[:,0], pcadf.loc[:,1], pcadf.loc[:,2],c = wine_df['quality'], cmap='viridis')
plt.show()
In [1]:
fig = plt.figure()
ax = fig.add_subplot(projection='3d')
ax.scatter(pcadf.loc[hilowines.index,0], pcadf.loc[hilowines.index,1], pcadf.loc[hilowines.index,2],c = hilowines['quality'], cmap='viridis')
plt.show()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-1-282003e198df> in <module>
----> 1 fig = plt.figure()
      2 ax = fig.add_subplot(projection='3d')
      3 ax.scatter(pcadf.loc[hilowines.index,0], pcadf.loc[hilowines.index,1], pcadf.loc[hilowines.index,2],c = hilowines['quality'], cmap='viridis')
      4 plt.show()
      5 

NameError: name 'plt' is not defined
In [26]:
pearsonr(wine_df['quality'], pcadf.loc[:,0])
Out[26]:
(0.11034929787171394, 9.749578010208338e-06)
In [27]:
pearsonr(wine_df['quality'], pcadf.loc[:,1])
Out[27]:
(-0.38680246906124865, 3.226863196143725e-58)
In [28]:
pearsonr(wine_df['quality'], pcadf.loc[:,2])
Out[28]:
(0.39927380944216867, 2.9790456332805242e-62)
In [29]:
pearsonr(wine_df['quality'], wine_df['alcohol'])
Out[29]:
(0.476166324001136, 2.831476974778582e-91)
In [31]:
plt.scatter(wine_df['quality'], wine_df['alcohol'])
Out[31]:
<matplotlib.collections.PathCollection at 0x211d9412730>
In [24]:
pcadf.corr()
Out[24]:
0 1 2 3 4
0 1.000000e+00 -1.684594e-17 -1.486133e-16 -2.173225e-16 -5.815789e-17
1 -1.684594e-17 1.000000e+00 2.351692e-16 -2.316097e-16 5.963830e-18
2 -1.486133e-16 2.351692e-16 1.000000e+00 -3.023205e-16 2.089346e-17
3 -2.173225e-16 -2.316097e-16 -3.023205e-16 1.000000e+00 -4.061744e-16
4 -5.815789e-17 5.963830e-18 2.089346e-17 -4.061744e-16 1.000000e+00
In [25]:
from scipy.stats import pearsonr
In [32]:
pearsonr(pcadf.loc[:,0], wine_df['quality'])
Out[32]:
(0.11034929787171394, 9.749578010208338e-06)
In [33]:
pearsonr(pcadf.loc[:,1], wine_df['quality'])
Out[33]:
(-0.38680246906124865, 3.226863196143725e-58)
In [34]:
pearsonr(pcadf.loc[:,2], wine_df['quality'])
Out[34]:
(0.39927380944216867, 2.9790456332805242e-62)
In [38]:
pcX = np.array(pcadf.loc[:,0:2])
In [44]:
Y = np.array(wine_df['quality'])
In [45]:
lr.fit(pcX, Y)
Out[45]:
LinearRegression()
In [46]:
lr.score(pcX, Y)
Out[46]:
0.32121269251911977
In [47]:
lr.coef_
Out[47]:
array([ 0.05060501, -0.22501699,  0.25886473])
In [42]:
from sklearn.linear_model import LinearRegression
lr = LinearRegression()
In [51]:
plt.scatter(pcadf.loc[:,1], pcadf.loc[:,3], c=wine_df['quality'], cmap='viridis')
plt.show()
In [ ]: