forked from cxybaba/MachineLearning
-
Notifications
You must be signed in to change notification settings - Fork 0
/
503indig2.py
175 lines (161 loc) · 6.54 KB
/
503indig2.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
#['BOD5_ET', 'COD_ET', 'TN_ET', 'TP_ET', 'PH_ET', 'NH4_N_ET']
from sklearn import linear_model
import numpy as np
import pandas as pd
import matplotlib.pylab as plt
import matplotlib.dates as mdates
from sklearn.decomposition import PCA
from sklearn.model_selection import train_test_split
from sklearn import svm
from sklearn.model_selection import ShuffleSplit
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import cross_val_predict
from sklearn.metrics import r2_score,mean_squared_error
from sklearn.cluster import KMeans
from sklearn import metrics
from sklearn.model_selection import KFold
from sklearn.tree import DecisionTreeRegressor
data = pd.read_csv('503afterwash2.csv',encoding='utf-8',index_col='YMD')#变成了dataframehahaxixi
data.index = pd.to_datetime(data.index)#这一列作为时间序列
#NH$从367开始有值
datanp = np.array(data)[367:587]
# print(datanp)
# datain,dataout = np.split(datanp,(6,), axis=1)#分割输入输出
# print(datain)
# print(dataout)
#ndarray变成list
# datainlist = datain.tolist()
# dataoutlist = dataout.tolist()
# print(datainlist)
# print(dataoutlist)
#将有值的部分重新组成datafram
data2 = pd.DataFrame(datanp,columns=['BOD5_ET', 'COD_ET', 'TN_ET', 'TP_ET', 'PH_ET', 'NH4_N_ET'])
print(data2.tail())
#***********************维数选择**************************** 填补NH4的结果不错
#相关性检验 结合相关性检验选择 DLQLTY BOD5_ET TN_ET TP_ET 作为输入
print(data2.corr(method='pearson')['NH4_N_ET'])
#pca检验需要的维数 效果不明显
# pca = PCA(copy=True, iterated_power='auto', n_components=6, random_state=None,
# svd_solver='randomized', tol=0.0, whiten=False)
# a,b = np.split(datanp,(6,), axis=1)
# print(a.tolist()) 把输入独立出来
# pca.fit(a)
# print(pca.explained_variance_ratio_) #输出主成分比例,也就是一维的时候，二维的时候，三维的时候
def har():
index = ['NH4_N_ET','TN_ET','TP_ET','COD_ET','BOD5_ET','']
data = np.array(data2.corr(method='pearson')['NH4_N_ET']).tolist()
# print(sorted(data))
plt.barh(bottom=(0,1,2,3,4,5),height=0.35,width=sorted(data)[::-1],align="center")
plt.yticks((0,1,2,3,4,5),index)
ax = plt.gca()
ax.spines['right'].set_color('none')#ax的spines指的就是图形的四个边框，脊梁
ax.spines['top'].set_color('none')
ax.yaxis.set_ticks_position('left')#将左边的边框作为y轴
ax.spines['left'].set_position(('data',0))#挪动边框 在哪个点落位，y轴在x等于-1落位
plt.text(0.08, 4.85, 'PH_ET', ha='center', va= 'bottom',fontsize=10)
plt.plot([0.3,0.3,0.3,0.3,0.3,0.3,0.3],[-1,0,1,2,3,4,5],color='red', linewidth=1.0, linestyle='--')
plt.text(0.32, 4.5, 'corr=0.3', ha='left', va= 'bottom',fontsize=10,color='red')
plt.show()
har()
datanp2 = np.array(data2.as_matrix(columns = ['BOD5_ET','COD_ET','TN_ET', 'TP_ET','NH4_N_ET']))
datain,dataout = np.split(datanp2,(4,), axis=1)#分割输入输出
#输入输出的list
k=220 #选择220条数据效果最好
datain1 = datain[:k]
dataout1 = dataout[:k]
clf = linear_model.BayesianRidge(alpha_1=1000, alpha_2=1000, compute_score=False,
copy_X=True, fit_intercept=True, lambda_1=1000, lambda_2=1000,
n_iter=3000, normalize=False, tol=1000, verbose=False)
clf1 = svm.SVR(kernel='linear', C=1,epsilon=5)#0.1 0.1 0.1就超越了贝叶斯 C影响较大，C越大运行越慢，epsilon越大运行越快 通过pso去优化C和epsilon的值 C和epsilon取在（0，5]优化
clf2 = DecisionTreeRegressor(max_depth=4)
def MAPE(true, pred):
diff = np.abs(np.array(true) - np.array(pred))
return np.mean(diff / true)
def testclf(clf):
cv = ShuffleSplit(n_splits=5, train_size=0.8, random_state=0)#train_size表示的是整个数据数据集选多少train来训练
scores = cross_val_score(clf, datain1, dataout1, cv=cv)
print(scores)
clf.fit(datain1,dataout1)
predicted = clf.predict(datain[:k])
#print(r2_score(predicted,dataout[:k]))
#检查预测结果和实际结果的相关性
b = []
for n in dataout[:k]:
b.append(n[0])
d = {
'predict':predicted.tolist(),
'true':b
}
datacor = pd.DataFrame(d)
print(datacor.corr(method='pearson'))
#print(MAPE(b,predicted.tolist()))
# print('beiyesi:','*****************')
# testclf(clf)
# print('jueceshu:','*****************')
# testclf(clf2)
clf2 = DecisionTreeRegressor(max_depth=3)
# print('svr:','*****************')
# testclf(clf1)
def kfoldtest(clf):
score = []
r2s = []
corrs = []
kf=KFold(5,shuffle=False,random_state=0)
x = datain[:k]
y = dataout[:k]
for train_index, test_index in kf.split(x,y):
X_train, X_test, y_train, y_test = x[train_index], x[test_index], y[train_index], y[test_index]
clf.fit(X_train,y_train)
predicted = clf.predict(X_test)
r2 = round(r2_score(predicted,y_test),3)
r2s.append(r2)
# print('r2_score2:',round(r2_score(predicted,y_test),3))
b = []
for i in y_test.tolist():
b.append(i[0])
d = {
'predict':predicted.tolist(),
'true':b
}
datacor = pd.DataFrame(d)
corr = round(datacor.corr(method='pearson')['predict'][1],5)
corrs.append(corr)
score.append(r2s)
score.append(corrs)
return score
# print('coor:',round(datacor.corr(method='pearson')['predict'][1],3))
print('svrkfold:','*****************')
print(np.array(kfoldtest(clf1)[0]),np.array(kfoldtest(clf1)[1]))
print('beiyesi:','*****************')
print(np.array(kfoldtest(clf)[0]).mean(),np.array(kfoldtest(clf)[1]).mean())
print('jueceshu:','*****************')
print(np.array(kfoldtest(clf2)[0]).mean(),np.array(kfoldtest(clf2)[1]).mean())
# print('svr',20*'*')
# clf = svm.SVR(kernel='rbf', C=1000)
# cv = ShuffleSplit(n_splits=5, test_size=0.2, random_state=0)
# scores = cross_val_score(clf, datain1, dataout1, cv=cv)
# print(scores)
# clf.fit(datain1,dataout1)
# predicted = clf.predict(datain[k:])
# # print(predicted)
# print(r2_score(predicted,dataout[k:]))
#**********************聚类分析*******************
# datanp3 = np.array(data2.as_matrix(columns = ['BOD5_ET', 'COD_ET', 'TN_ET', 'TP_ET', 'PH_ET']))
# # print(datanp3.tolist())
# #Kmeans
# clf = KMeans(n_clusters=8,algorithm='full',max_iter=1000)
# km = clf.fit(datanp3.tolist())
# print(km.labels_)
# #评价聚类模型 分成四维
# a,b= np.split(datanp3,(258,), axis=0)#分割输入输出
# def findharabaz_score():
# harabaz_score = []
# k = [x for x in range(4,121)]
# for x in k:
# clf = KMeans(n_clusters=x,algorithm='full',max_iter=1000)
# km = clf.fit(a)
# harabaz_score.append(metrics.calinski_harabaz_score(a,km.labels_))
# plt.figure()
# plt.plot(k,harabaz_score)#往figure里面装数据
# plt.show()
# # findharabaz_score() #分成8维