whiskies.py

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

whisky = pd.read_csv('whiskies.txt')
whisky['region'] = pd.read_csv('regions.txt')

flavors = whisky.iloc[:, 2:14]

corr_flavors = pd.DataFrame.corr(flavors)

plt.figure(figsize=(10,10))
plt.pcolor(corr_flavors)
plt.colorbar()
plt.savefig('corr_flavors.jpg')

corr_whisky = pd.DataFrame.corr(flavors.transpose())
plt.figure(figsize=(10,10))
plt.pcolor(corr_whisky)
plt.colorbar()
plt.savefig('corr_whisky.jpg')

import sklearn.cluster.bicluster

model = sklearn.cluster.bicluster.SpectralCoclustering(n_clusters=6, random_state=0)
model.fit(corr_whisky)
np.sum(model.rows_, axis=1)

whisky['Group'] = pd.Series(model.row_labels_, index=whisky.index)
whisky = whisky.iloc[np.argsort(model.row_labels_)]
whisky = whisky.reset_index(drop=True)

correlations = pd.DataFrame.corr(whisky.iloc[:,2:14].transpose())
correlations = np.array(correlations)

plt.figure(figsize=(14,7))
plt.subplot(121)
plt.pcolor(corr_whisky)
plt.title('Original')
plt.axis('tight')
plt.subplot(122)
plt.pcolor(correlations)
plt.title('Rearranged')
plt.axis('tight')
plt.savefig('correlations.jpg')