import pandas as pd
import sklearn.metrics as met
from sklearn import preprocessing
from sklearn.cluster import AgglomerativeClustering
import matplotlib.pyplot as plt

import subprocess

def make_tree(children, instances, labels, file_name ):

    with open('dt.dot', 'w') as f:
        f.write("""
      digraph G {
    nodesep=0.3;
    ranksep=0.2;
    margin=0.1;
    node [shape=circle];
    edge [arrowsize=0.8];

        """)
        labels=labels.tolist()
        for i in range(0, len(labels)):
            f.write(str(i) + '[label="'+ labels[i] + '" color=blue];\n')

        for x in children:
            f.write(str(instances) + "->" + str(x[0].item()) + ";\n")
            f.write(str(instances) + "->" + str(x[1].item()) + ";\n")
            instances+=1

        f.write("}")
        f.close()

        subprocess.call("dot -Tpng dt.dot -o " + file_name + ".png", shell=True)


df = pd.read_csv("dogs.csv")

#prikaz imena kolona + 5 prvih instanci
print('Prvih 5 instanci', df.head(), sep='\n')
print('\n\n')

num_instances = df.shape[0]
featurs = df.columns[1:3].tolist()

#print(featurs)
x_original=df[featurs]

#standardizacija atributa
x=pd.DataFrame(preprocessing.scale(x_original))

#normalizacija
#x=pd.DataFrame(preprocessing.MinMaxScaler().fit_transform(x_original))

#dodeljivanje imena kolonama
x.columns = featurs

"""
AgglomerativeClustering

parametri:

n_clusters : broj klastera
             default=2
connectivity : matrica povezanosti
               default=None
affinity : mera za racnunanje bliskosti
          default: “euclidean”
          Moze biti npr. “euclidean”, “l1”, “l2”, “manhattan”, “cosine”
linkage : veza
          default: “ward”
          Moze biti “ward”, “complete”, “average”

atributi:
labels_ : oznake klastera za svaku instancu
n_leaves_ : broj listova u hijerahijskom drvetu
children_ : deca cvorova koji nisu listovi. Vrednosti manje od broja instanci u skupu predstavljaju indekse instanci.
"""

colors = ['red', 'green', 'gold', 'blue', 'black']
fig = plt.figure()
plt_ind=1

for i in range(2,5):
    for link in ['complete', 'average']:
        est=AgglomerativeClustering(n_clusters=i, linkage=link)
        est.fit(x)
        make_tree(est.children_, num_instances, df['breed'], 'clus_' + str(i) + '_' + link )

        df['labels']= est.labels_
        #print('Prvih 5 instanci', df.head(), sep='\n')

        fig.add_subplot(3, 2, plt_ind)
        for j in range(0,i):
            cluster= df.loc[lambda x: x['labels'] == j, :]
            plt.scatter(cluster['height'], cluster['weight'], color=colors[j], s=30, marker='o', label="cluster %d"%j)

        plt.legend(loc='lower right')
        plt.title('Broj klastera: %d, veza: %s' % (i, link))
        plt_ind += 1

plt.tight_layout()
plt.show()