-
Notifications
You must be signed in to change notification settings - Fork 1
/
Cluster.py
206 lines (184 loc) · 7.87 KB
/
Cluster.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
import pandas as pd
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
from sklearn import preprocessing
import numpy as np
from analyze.reorgdata import callAllfile
from sklearn.preprocessing import OneHotEncoder
import json
import scipy.stats as sci
# feature:[lat,lon,weekday_checkin,weekday_checkout,weekend_checkin,weekend_checkout]
# k={20，30，40]
# output: cluster{station_id} to csv
# average check-in / check-out demand for stations in workday or holiday
def diffTwolist(l1,l2):
return list(set(l1)-set(l2))
def avgDemand(demand_type,path,start_date, end_date):
# demand_type: Start / End
# path: 'D:\my\dataset\citibike\data\day\demand\' /// 'D:\my\dataset\captitalbike\data\day\demand\'
# output: (work_day demand and holiday_demand) for demand_type
demand_f=path+'day/demand/'
if demand_type=='Start':
demand_f = demand_f+'demand_eaStart.csv'
if demand_type=='End':
demand_f = demand_f+'demand_eaEnd.csv'
demand_df = pd.read_csv(demand_f)
calendar = pd.read_csv('D:\my\dataset\calendar\calendar.csv')
# find workday
if end_date=='':
work_day = \
calendar[(calendar['workday'] == 1) & (calendar['Date'] > start_date)][
'Date'].values.tolist()
holi_day = \
calendar[(calendar['workday'] == 0) & (calendar['Date'] > start_date)][
'Date'].values.tolist()
else:
work_day = calendar[(calendar['workday']==1)&(calendar['Date']>start_date)&(calendar['Date']<end_date)]['Date'].values.tolist()
holi_day = calendar[(calendar['workday']==0)&(calendar['Date']>start_date)&(calendar['Date']<end_date)]['Date'].values.tolist()
num_wd = len(work_day)
num_holi = len(holi_day)
# print(work_day)
# no demands in some days
diff_set4work = diffTwolist(work_day,demand_df.columns)
for d in diff_set4work:
work_day.remove(d)
diff_set4holi = diffTwolist(holi_day,demand_df.columns)
for d in diff_set4holi:
holi_day.remove(d)
work_demand = demand_df[work_day]
holi_demand=demand_df[holi_day]
demand4starions_workday = work_demand.sum(axis=1)/num_wd
demand4station_holiday = holi_demand.sum(axis=1)/num_holi
demand_df = pd.concat([demand4starions_workday,demand4station_holiday],axis=1)
col1 = demand_type+'_'+'workday'
col2 = demand_type+'_'+'holiday'
# change col name
demand_df.rename(columns = {0:col1,1:col2},inplace=True)
return demand_df
def NORMARLIZATION(df):
feat = df.values
mm_scaler = preprocessing.MinMaxScaler()
feat_sacler = mm_scaler.fit_transform(feat)
return feat_sacler
def cluster(cluster_n,path,start_date,end_date):
# start_date : citibike:(2013-06-01), capitalbike()
# end_date: citibike(2017-11-30), captitalbike()
# return station sets
kwds='index_num'
# if 'capi' in path:
# kwds = 'short_name'
# else:
# kwds = 'station_id'
station_infor_path = path+'station_info.csv'
station_info_df = pd.read_csv(station_infor_path)
Start_demand = avgDemand('Start',path,start_date,end_date)
End_demand = avgDemand('End',path,start_date,end_date)
feature_df = pd.concat([Start_demand,End_demand],axis=1)
feature_df = pd.concat([station_info_df[[kwds,'lat','lon']],feature_df],axis=1)
# drop the rows without demand
# real_bike = feature_df[(feature_df['Start_holiday']!=0)&(feature_df['Start_workday']!=0)&(feature_df['End_workday']!=0)&(feature_df['End_holiday']!=0)]
real_bike = feature_df
real_bike_feature = real_bike[['lat','lon','Start_holiday','Start_workday','End_holiday','End_workday']]
nm_feat = NORMARLIZATION(real_bike_feature)
C_MODEL = KMeans(cluster_n)
C_MODEL.fit(nm_feat)
label = C_MODEL.labels_
center = C_MODEL.cluster_centers_
# print(len(real_bike_feature['lat'].values),len(real_bike_feature['lon'].values))
# real_bike_feature.plot.scatter(x='lat',y='lon',color='b')
plt.scatter(nm_feat[:,0],nm_feat[:,1],c=label)
plt.scatter(center[:,0],center[:,1],c='r',marker='s')
plt.show()
cluster_r = np.concatenate((np.reshape(real_bike[kwds].values,newshape=[-1,1]),np.reshape(label,newshape=[-1,1])),axis=1)
# print(cluster_r)
stationToonehotcluster(path,cluster_r,cluster_n)
return cluster_r
def stationToonehotcluster(path, cluster_list,cluster_n):
dic = {}
# {label:[station_id]}
for station_id, label in cluster_list:
# print(station_id,label)
if station_id not in dic.keys():
dic[int(station_id)] = int(label)
station_f = path + 'station_info.csv'
station_df = pd.read_csv(station_f)
station_c = []
stations = station_df['index_num'].values
onehot = OneHotEncoder(sparse=False)
for i in stations:
c_i = dic[i]
station_c.append(c_i)
# save station_id + cluster_id into a file
j = json.dumps(dic)
f_n =path+str(cluster_n)+ "scid.json"
w = open(f_n,'w')
w.write(j)
w.close()
station_c = np.expand_dims(station_c, -1)
station_c_one = onehot.fit_transform(station_c)
sc_f = path + str(cluster_n)+'station_cluster.csv'
np.savetxt(sc_f, station_c_one, delimiter=',', newline='\n')
def calProb4eacluster(path,cluster_n,dur_stop):
# probability of one cluster to others
# cluster_list [station_id,label]
# print(cluster_list)
dic = json.load(open('D:\my\dataset\citibike\data\\20scid.json','r'))
c_metric = np.zeros(shape=[cluster_n,cluster_n])
# find edges and fill the c_metric (2010-9-20,2011-9-20)
dir = path+'day\graph/'
file_list = callAllfile(dir)
i=0
for file in file_list:
# Daily cluster distribution
c_daily_m = np.zeros(shape=[cluster_n, cluster_n])
if i==dur_stop:
break
df = pd.read_csv(file)
for index,row in df.iterrows():
# print(index)
if str(row['source'].astype(int)) not in dic.keys() or str(row['target'].astype(int)) not in dic.keys(): continue
c_s = dic[str(row['source'].astype(int))]
c_t = dic[str(row['target'].astype(int))]
c_metric[c_s,c_t]+=1
c_metric[c_t,c_s]+=1
c_daily_m[c_t,c_s]+=1
c_daily_m[c_s,c_t]+=1
i+=1
c_daily_sum = np.sum(c_daily_m,axis=1)
c_daily_m = c_daily_m/c_daily_sum
c_daily_m[np.isnan(c_daily_m)]=0.0
f_d = file.split('\\')[-1]
c_daily_f = path + 'day\graph\cluster_graph\\'+f_d
np.savetxt(c_daily_f,c_daily_m,delimiter=',',newline='\n')
c_sum = np.sum(c_metric,axis=1)
c_prob = np.zeros_like(c_metric)
c_prob = c_metric/c_sum
c_prob[np.isnan(c_prob)]=0.0
c_metric_f = path+'cluster.csv'
# print(c_prob)
np.savetxt(c_metric_f,c_prob,delimiter=',',newline='\n')
# station_c_pro = np.matmul(station_c_one,c_prob)
# station_c_pro_f = path + 'station_c_pro.csv'
# np.savetxt(station_c_pro_f,X=station_c_pro,delimiter=',')
def corrGraph(path,start_date,end_date):
kwds = 'index_num'
station_infor_path = path+'station_info.csv'
station_info_df = pd.read_csv(station_infor_path)
Start_demand = avgDemand('Start',path,start_date,end_date)
End_demand = avgDemand('End',path,start_date,end_date)
feature_df = pd.concat([Start_demand,End_demand],axis=1)
feature_df = pd.concat([station_info_df[[kwds,'lat','lon']],feature_df],axis=1)
real_bike_feature = feature_df[['Start_holiday','Start_workday','End_holiday','End_workday']]
nm_feat = NORMARLIZATION(real_bike_feature)
T_feat= pd.DataFrame(data=nm_feat,columns=['Start_holiday','Start_workday','End_holiday','End_workday']).T
sToSpearson = T_feat.corr(method='pearson')
st_corr_f = path +'sTos_corr.csv'
sToSpearson.to_csv(st_corr_f)
#
path = 'D:\my\dataset\citibike\data/'
start_date = '2016-10-01'
end_date = '2017-10-28'
cluster_n = 20
# cluster_res = cluster(cluster_n,path,start_date,end_date)
# calProb4eacluster(path,cluster_n,389)
# corrGraph(path,start_date,end_date)