import scipy

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

import seaborn as sns

from sklearn import preprocessing, linear_model, svm

from sklearn.metrics import f1_score, accuracy_score

from sklearn.model_selection import train_test_split

from sklearn import metrics

from sklearn.metrics import confusion_matrix, classification_report

from sklearn.metrics.cluster import normalized_mutual_info_score

from sklearn import cluster

from sklearn.preprocessing import StandardScaler

from sklearn.decomposition import PCA

from sklearn import manifold

from sklearn.neighbors import NearestNeighbors

#convert csv of data set to pandas dataframe`

df = pd.read_csv("winequality-white.csv", sep=';')

df.head()

df.describe()

corr = df.corr()

#plot correlation heatmap between all columns in dataframe

plt.subplots(figsize=(15,10))

ax = plt.axes()

ax.set_title("Wine Characteristic Correlation Heatmap")

sns.heatmap(corr,

xticklabels=corr.columns,

yticklabels=corr.columns,

annot=True,

cmap=sns.diverging_palette(220, 20, as_cmap=True))

#histogram of quality

fig = plt.figure(figsize = (8,8))

plt.hist(df.quality,bins=6,alpha=0.5,histtype='bar',ec='black')

plt.title('Distribution of the Quality')

plt.xlabel('Quality')

plt.ylabel('Count')

#boxplot of quality and alcohol

plt.figure(figsize=(8,5))

sns.boxplot(x='quality',y='alcohol',data=df,palette='GnBu_d')

plt.title("Boxplot of Quality and Alcohol")

plt.show()

cols = ["total sulfur dioxide","free sulfur dioxide","residual sugar","fixed acidity","volatile acidity","alcohol","sulphates","pH","density", "citric acid", "chlorides"]

#separate data into training and test sets and choose state so that results are repeatable

X_train, X_test, y_train, y_test = train_test_split(df[cols], df["quality"], test_size=0.2, random_state=4)

#declare and fit svm with rbf kernel to training set

clf = svm.SVC(kernel="rbf", gamma=1, C=1,

decision_function_shape="ovo")

clf.fit(X_train,y_train)

ypred = clf.predict(X_test)

#get scores for how well the svm performs on test set

print("Accuracy rbf kernel: %.2f"

% clf.score(X_test, y_test))

print("nmi rbf kernel: %.2f"

% normalized_mutual_info_score(y_test, ypred))

print(classification_report(y_test, ypred))

X = StandardScaler().fit_transform(df[cols])

#finding ideal eps for DBSCAN by calculating the distance to the nearest n points for each point,

#sorting and plotting the results. Then we look to see where the change is most pronounce

#and select that as epsilon.

neigh = NearestNeighbors(n_neighbors=2)

nbrs = neigh.fit(X)

distances, indices = nbrs.kneighbors(X)

distances = np.sort(distances, axis=0)

distances = distances[:,1]

plt.plot(distances)

plt.title("finding optimal eps value for DBSCAN using elbow method")

params = {'n_clusters': 2, 'eps': 2, 'min_samples': 5}

spectral = cluster.SpectralClustering(n_clusters=params['n_clusters'],

eigen_solver='arpack',

affinity="nearest_neighbors")

dbscan = cluster.DBSCAN(eps=params['eps'],

min_samples=params['min_samples'])

clustering_algorithms = (

('SpectralClustering', spectral),

('DBSCAN', dbscan),

)

y_pred = {}

for name, algorithm in clustering_algorithms:

y_pred[name] = algorithm.fit_predict(X)

#calculate number of true labels between 1-9 in each cluster and then choose the label which appears

#most frequently as the label of that cluster

avg_quality_spec = [[0, 0, 0, 0, 0, 0, 0, 0 ,0], [0, 0, 0, 0, 0, 0, 0]]

avg_quality_dbscan = [[0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0]]

for i, pred in enumerate(y_pred['DBSCAN']):

if pred == 0:

avg_quality_dbscan[0][df['quality'].iloc[i] - 3] += 1

elif pred == -1:

avg_quality_dbscan[1][df['quality'].iloc[i] - 3] += 1

for i, pred in enumerate(y_pred['SpectralClustering']):

if pred == 0:

avg_quality_spec[0][df['quality'].iloc[i] - 3] += 1

elif pred == 1:

avg_quality_spec[1][df['quality'].iloc[i] - 3] += 1

# in all the clusters, the quality of wine that appears the most in 6 so all elements

#in dataframe are labeled as 6

y_pred_ = [6 for i in range(4898)]

print("f1_score: %0.3f"

% f1_score(df['quality'], y_pred_, average='weighted'))

print("accuracy_score: %.03f"

% accuracy_score(df['quality'], y_pred_))

print("Mutual Information: %0.3f"

% metrics.normalized_mutual_info_score(df['quality'], y_pred_))

#for labels in y_pred.values():

# print("Homogeneity: %0.3f" % metrics.homogeneity_score(labels_true, labels))

# print("Completeness: %0.3f" % metrics.completeness_score(labels_true, labels))

# print("V-measure: %0.3f" % metrics.v_measure_score(labels_true, labels))

# print("Adjusted Rand Index: %0.3f"

# % metrics.adjusted_rand_score(labels_true, labels))

# print("Silhouette Coefficient: %0.3f"

# % metrics.silhouette_score(X, labels))

#dimension reduction with pca

pca = PCA(n_components=3).fit(X)

new_cor = pca.transform(X)

#plot projection of 3d pca onto two components

fig = plt.figure(figsize=(15,10))

fig.add_subplot(1,3,1)

plt.scatter(new_cor[:,0], new_cor[:,1], marker='x', cmap='jet')

plt.xlabel("PCA Component 1")

plt.ylabel("PCA Component 2")

fig.add_subplot(1,3,2)

plt.scatter(new_cor[:,0], new_cor[:,2], marker='x', cmap='jet')

plt.xlabel("PCA Component 1")

plt.ylabel("PCA Component 3")

fig.add_subplot(1,3,3)

plt.scatter(new_cor[:,1], new_cor[:,2], marker='x', cmap='jet')

plt.xlabel("PCA Component 2")

plt.ylabel("PCA Component 3")

#pearson correlation coefficient for all the pca components with quality

corr1 = scipy.stats.pearsonr(new_cor[:,0], df['quality'])

corr2 = scipy.stats.pearsonr(new_cor[:,1], df['quality'])

corr3 = scipy.stats.pearsonr(new_cor[:,2], df['quality'])

print(corr1)

print(corr2)

print(corr3)

#plot information content of each pca component

fig = plt.figure(figsize=(12,6))

plt.bar(np.arange(pca.n_components_), 100*pca.explained_variance_ratio_)

plt.title('Relative information content of PCA components')

plt.xlabel("PCA component number")

plt.ylabel("PCA component variance % ")

#dimension reduction with Isomap

iso = manifold.Isomap(n_neighbors=6, n_components=3).fit(X)

manifold_3Da = iso.transform(X)

#plot projection of 3d isomap onto two components

fig = plt.figure(figsize=(15,10))

fig.add_subplot(1,3,1)

plt.scatter(manifold_3Da[:,0], manifold_3Da[:,1], marker='x', cmap='jet')

plt.xlabel("Isomap Component 1")

plt.ylabel("Isomap Component 2")

fig.add_subplot(1,3,2)

plt.scatter(manifold_3Da[:,0], manifold_3Da[:,2], marker='x', cmap='jet')

plt.xlabel("Isomap Component 1")

plt.ylabel("PCA Component 3")

fig.add_subplot(1,3,3)

plt.scatter(manifold_3Da[:,1], manifold_3Da[:,2], marker='x', cmap='jet')

plt.xlabel("Isomap Component 2")

plt.ylabel("Isomap Component 3")

plt.show()

#pearson correlation coefficient for all the Isomap components with quality

corr1 = scipy.stats.pearsonr(manifold_3Da[:,0], df['quality'])

corr2 = scipy.stats.pearsonr(manifold_3Da[:,1], df['quality'])

corr3 = scipy.stats.pearsonr(manifold_3Da[:,2], df['quality'])

print(corr1)

print(corr2)

print(corr3)