import numpy as np
import pandas as pd
import csv, os

import matplotlib.pyplot as plt
import matplotlib
matplotlib.style.use('ggplot')
%matplotlib inline

styles = pd.read_excel('LOC_NaughtyBoy/abv_ibu_comp.xlsx')
styles.rename(columns={'nonsense':'style'}, inplace=True)
print styles.shape
print styles.head()
print styles['style'].value_counts()
styles.describe()

styles_noNAN = styles.dropna()

print styles_noNAN.shape
print styles_noNAN.head()
print styles_noNAN['style'].value_counts()
styles_noNAN.describe()

styles_noNAN.plot(kind='scatter', x='abv', y='ibu')

bipa = styles_noNAN['style'] == 'bipa'
ipa = styles_noNAN['style'] == 'ipa'
stout = styles_noNAN['style'] == 'stout'

plt.figure(figsize=(10,6))

plt.scatter(styles_noNAN[bipa]['abv'], styles_noNAN[bipa]['ibu'], c='brown', edgecolor='k', marker='D', s=50, alpha=.7, label='BIPA')
plt.scatter(styles_noNAN[ipa]['abv'], styles_noNAN[ipa]['ibu'], c='yellow', marker='o', s=50, alpha=.7, label='IPA')
plt.scatter(styles_noNAN[stout]['abv'], styles_noNAN[stout]['ibu'], c='black', marker='s', s=50, alpha=.7, label='Stout')

plt.title('NL')
plt.xlabel('abv')
plt.ylabel('ibu')

plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)

def distance(p0, p1):
    # squared Euclidean distance
    return np.sum( (p0-p1)**2 )

def nn_classify(training_set, training_labels, new_example):
    dists = np.array([distance(t, new_example) for t in training_set])
    nearest = dists.argmin()
    return training_labels[nearest]

beers = np.array(styles_noNAN[['abv', 'ibu']])
labels = np.array(styles_noNAN['style'])

print beers[:5]
print labels[:5]

nn_classify(beers, labels, [8., 60.])

yIBU = np.linspace(0, 120)
xABV = np.linspace(2, 14)

xyGrid = []
xyLab = []

for x in np.nditer(xABV):
    for y in np.nditer(yIBU):
        xyGrid.append([x, y])
        label = nn_classify(beers, labels, (x,y))
        xyLab.append(label)

xyGrid = np.array(xyGrid)
xyLab = np.array(xyLab)
print len(xyGrid)
print len(xyLab)
print xyLab[:10]

colors = {a:b for a,b in zip(np.unique(labels), ['brown', 'yellow', 'black'])}
print colors

plt.figure(figsize=(10,6))

for dot in range(len(xyGrid)):
    plt.plot(xyGrid[dot,0], xyGrid[dot,1], marker='s', ms=12, alpha=.3, c=colors[xyLab[dot]])

plt.figure(figsize=(10,6))

for dot in range(len(xyGrid)):
    plt.plot(xyGrid[dot,0], xyGrid[dot,1], marker='s', ms=12, alpha=.3, c=colors[xyLab[dot]])

plt.scatter(styles_noNAN[bipa]['abv'], styles_noNAN[bipa]['ibu'], c='brown', edgecolor='k', lw=2, marker='D', s=50, alpha=.9, label='BIPA')
plt.scatter(styles_noNAN[ipa]['abv'], styles_noNAN[ipa]['ibu'], c='yellow', edgecolor='k', lw=2, marker='o', s=50, alpha=.9, label='IPA')
plt.scatter(styles_noNAN[stout]['abv'], styles_noNAN[stout]['ibu'], c='black', edgecolor='w', lw=2, marker='s', s=50, alpha=.9, label='Stout')

plt.title('NL')
plt.xlabel('abv')
plt.ylabel('ibu')

plt.ylim(0, 120)
plt.xlim(2, 14)

plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)

# subtract the mean for each feature:
features = beers.copy()
features -= features.mean(axis=0)
# divide each feature by its standard deviation
features /= features.std(axis=0)

markers = {a:b for a,b in zip(np.unique(labels), ['D', 'o', 's'])}
colors = {a:b for a,b in zip(np.unique(labels), ['brown', 'yellow', 'black'])}

plt.figure(figsize=(10,6))

for beer in markers:
    current_beer = labels == beer
    plt.scatter(features[current_beer][:,0], features[current_beer][:,1], 
                marker=markers[beer], c=colors[beer], edgecolor='k', s=50, alpha=.7, label=beer
               )
    
plt.xlabel('abv')
plt.ylabel('ibu')

# plt.ylim((.75, 1.))

plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)

yIBU2 = np.linspace(-3, 3)
xABV2 = np.linspace(-4, 4)

xyGrid2 = []
xyLab2 = []

for x in np.nditer(xABV2):
    for y in np.nditer(yIBU2):
        xyGrid2.append([x, y])
        label = nn_classify(features, labels, (x,y))
        xyLab2.append(label)

xyGrid2 = np.array(xyGrid2)
xyLab2 = np.array(xyLab2)
print len(xyGrid2)
print len(xyLab2)
print xyLab2[:10]

markers = {a:b for a,b in zip(np.unique(labels), ['D', 'o', 's'])}
colors = {a:b for a,b in zip(np.unique(labels), ['brown', 'yellow', 'black'])}

plt.figure(figsize=(12,9))

for dot in range(len(xyGrid2)):
    plt.plot(xyGrid2[dot,0], xyGrid2[dot,1], marker='s', ms=15, alpha=.3, c=colors[xyLab2[dot]])

for beer in markers:
    current_beer = labels == beer
    plt.scatter(features[current_beer][:,0], features[current_beer][:,1], 
                marker=markers[beer], c=colors[beer], edgecolor='k', lw=2, s=50, alpha=.7, label=beer
               )

plt.title('NL')
plt.xlabel('abv (standardized)')
plt.ylabel('ibu (standardized)')
plt.ylim(-3, 3)
plt.xlim(-4, 4)

plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)

from sklearn.cross_validation import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn import metrics

print features[:5]
print labels[:5]

X = features
y = labels

X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=8)

# check classification accuracy of KNN with K=5
knn = KNeighborsClassifier(n_neighbors=5)
knn.fit(X_train, y_train)
y_pred = knn.predict(X_test)
print metrics.accuracy_score(y_test, y_pred)

from sklearn.cross_validation import cross_val_score

# 10-fold cross-validation with K=5 for KNN (the n_neighbors parameter)
knn = KNeighborsClassifier(n_neighbors=5)
scores = cross_val_score(knn, X, y, cv=10, scoring='accuracy')
print scores

# use average accuracy as an estimate of out-of-sample accuracy
print scores.mean()

# search for an optimal value of K for KNN
k_range = range(1, 31)
k_scores = []
for k in k_range:
    knn = KNeighborsClassifier(n_neighbors=k)
    scores = cross_val_score(knn, X, y, cv=10, scoring='accuracy')
    k_scores.append(scores.mean())
print k_scores

# plot the value of K for KNN (x-axis) versus the cross-validated accuracy (y-axis)
plt.plot(k_range, k_scores)
plt.xlabel('Value of K for KNN')
plt.ylabel('Cross-Validated Accuracy')

# 10-fold cross-validation with the best KNN model
knn = KNeighborsClassifier(n_neighbors=2)
print cross_val_score(knn, X, y, cv=10, scoring='accuracy').mean()

print '\tabv\t\tibu'
# subtract the mean for each feature:
features_revisit = beers.copy()
features_mean = features_revisit.mean(axis=0)
print features_mean
features_revisit -= features_mean
# divide each feature by its standard deviation
features_std = features_revisit.std(axis=0)
print features_std
features_revisit /= features_std

# naughty boy ABV
nb_abv = 4.9
nb_abv_stand = (nb_abv - features_mean[0])/features_std[0]
print nb_abv_stand

nb_standardized = np.ones([50,2])
nb_standardized[:,0] = nb_standardized[:,0] * nb_abv_stand
nb_standardized[:,1] = nb_standardized[:,1] * yIBU2

print nb_standardized[:5]

nb_standardized

markers = {a:b for a,b in zip(np.unique(labels), ['D', 'o', 's'])}
colors = {a:b for a,b in zip(np.unique(labels), ['brown', 'yellow', 'black'])}

plt.figure(figsize=(12,9))

for dot in range(len(xyGrid2)):
    plt.plot(xyGrid2[dot,0], xyGrid2[dot,1], marker='s', ms=15, alpha=.3, c=colors[xyLab2[dot]])

for beer in markers:
    current_beer = labels == beer
    plt.scatter(features[current_beer][:,0], features[current_beer][:,1], 
                marker=markers[beer], c=colors[beer], edgecolor='k', lw=2, s=50, alpha=.7, label=beer
               )
    
plt.plot(nb_standardized[:,0], nb_standardized[:,1], 
         marker='o', ms=15, alpha=.3, c='green', label='naughty boy?')

plt.title('Naughty Boy v. Three Styles of Beers')
plt.xlabel('abv (standardized)')
plt.ylabel('ibu (standardized)')
plt.ylim(-3, 3)
plt.xlim(-4, 4)

plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)

neigh = KNeighborsClassifier(n_neighbors=2)
neigh.fit(X, y)

print neigh.predict(nb_standardized)
print ''
style_pred, count_pred = np.unique(neigh.predict(nb_standardized), return_counts=True)
print style_pred, count_pred

np.hstack((neigh.predict_proba(nb_standardized), neigh.predict(nb_standardized).reshape(50,1), nb_standardized))

# np.hstack((neigh.predict_proba(nb_standardized), nb_standardized))
nb_standardized[neigh.predict_proba(nb_standardized)[:,2] > 0]

# if NB has an IBU of 55 or greater, perhaps consider recategorizing to BIPA
stout_ibus = [-1.65306122, -0.30612245]
for ibu in stout_ibus:
    print ibu * features_std[1] + features_mean[1]
    
print "\n55 IBU standardized: ", (55. - features_mean[1])/features_std[1] 

nb_style = np.hstack((neigh.predict_proba(nb_standardized), neigh.predict(nb_standardized).reshape(50,1), nb_standardized))
nb_style[:5]

markers = {a:b for a,b in zip(np.unique(labels), ['D', 'o', 's'])}
colors = {a:b for a,b in zip(np.unique(labels), ['brown', 'yellow', 'black'])}

plt.figure(figsize=(12,9))

for dot in range(len(xyGrid2)):
    plt.plot(xyGrid2[dot,0], xyGrid2[dot,1], marker='s', ms=15, alpha=.3, c=colors[xyLab2[dot]])

for beer in markers:
    current_beer = labels == beer
    plt.scatter(features[current_beer][:,0], features[current_beer][:,1], 
                marker=markers[beer], c=colors[beer], edgecolor='k', lw=2, s=50, alpha=.7, label=beer
               )
    
# plt.plot(nb_standardized[:,0], nb_standardized[:,1], 
#          marker='o', ms=15, alpha=.3, c='green', label='naughty boy?')

for nb in nb_style:
    abv = nb[4]
    ibu = nb[5]
    color_list = []
    for prob in range(3):
        if nb[prob] > 0:
            color_list.append(colors[np.unique(labels)[prob]])
    plt.plot(abv, ibu, marker='o', ms=15, c=color_list[0], 
             markerfacecoloralt=color_list[-1],
             fillstyle='left',
#              label='naught boy!'            
            )

plt.plot(xABV2, np.ones(50)* -0.271265072934, 'w-', linewidth=3)
plt.text(-3.5, -0.2, '55 IBU', color='w', fontsize=25)

plt.title('Naughty Boy v. Three Styles of Beers')
plt.xlabel('abv (standardized)')
plt.ylabel('ibu (standardized)')
plt.ylim(-3, 3)
plt.xlim(-4, 4)

plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)