import pandas as pd
import sklearn
from sklearn.cluster import KMeans
from sklearn import preprocessing

le = preprocessing.LabelEncoder()



pd.set_option('display.max_columns', None)
#df1 = pd.read_excel('BVWEK33GZ17021703_MBB_Data Scientist Event.XLSX')
df1 = pd.read_excel('c0.xlsx')
le.fit(df1['vin'].unique())
df1['num_vin'] = le.transform(df1['vin'])
#print(df1.head())

#1 if successful, 0 if not
success_codes = ['202', '200', '200 - OK', '202 - Accepted']
df1['status_success'] = df1['responseStatus'].isin(success_codes)
df1['strange_browser'] = df1['userAgent'].isin(['Opera/9.80 Presto/2.10.x',
                        'Apache-HttpClient/4.5.2 (Java 1.5 minimum; Java/1.8.0_152)',
                        'Apache-HttpClient/4.5.2 (Java 1.5 minimum; Java/1.8.0_162)',
                        'Opera/9.80 Presto/2.10.x', 'Java/1.8.0_162'])



traceId_unique = df1['traceId'].unique()
parenttraceId_unique = df1['parentTraceId'].unique()
commonIdSet = set(traceId_unique) & set(parenttraceId_unique)
df1['hasParentId'] = df1['traceId'].isin(commonIdSet)
tIdsWithParent = df1[(df1['hasParentId'] == True)]


#clustering_columns = [df1['status_success'], df1['strange_browser'], df1['hasParentId'] ]


#k_means = df1[['status_success','strange_browser','hasParentId', 'processingTime', 'num_vin']]


# talk about this model! :::: ones: 127346, zeroes: 43972

scaler = preprocessing.MinMaxScaler()
scaler.fit(df1['processingTime'])

print('i fit')
#df1['scaled_times'] = scaler.fit_transform(df1['processingTime'])
df1['scaled_times'] = (df1.processingTime - df1.processingTime.min() ) / df1.processingTime.max() - df1.processingTime.min()
k_means = df1[['status_success','strange_browser', 'hasParentId',  'num_vin', 'scaled_times']]


print(k_means.head())

kmeans = KMeans(n_clusters=2, random_state=0).fit(k_means)

ones = [a for a in kmeans.labels_ if a == 1]
zeroes = [a for a in kmeans.labels_ if a == 0]

print(f'ones: {len(ones)}, zeroes: {len(zeroes)}')