-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathAll Variable BDT (Global Training)
192 lines (170 loc) · 6.14 KB
/
All Variable BDT (Global Training)
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
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import xgboost as xgb
from numpy import loadtxt
from xgboost import XGBClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import roc_curve, auc, accuracy_score
from sklearn.metrics import mean_squared_error as MSE
from sklearn.metrics import r2_score as r2
import pickle
csvpath = 'L1T_Jets_MLInputs_2018_SingleMu_PFA2.csv' #This file is available in the following link
# https://github.com/siddhesh86/Validation/blob/110X_hcalPUSub_dev/data/MLInputs/L1T_Jets_MLInputs_2018_SingleMu_PFA2.csv
data = pd.read_csv(csvpath, header = 0)
test_size = 0.50
randInt = 0
trainData, testData = train_test_split(data, random_state=randInt,
test_size=test_size, shuffle=False)
trainVars = data.columns.to_list()
trainVars = [x for x in trainVars if x not in ['PFJetEtCorr', 'L1JetType']]
# 'L1TJetDefault_PUS','L1JetDefault_PU','L1Jet9x9_RawEt','L1Jet9x9_EtSum7PUTowers','L1Jet7x9_RawEt','L1Jet7x9_EtSum7PUTowers','L1Jet5x9_RawEt','L1Jet5x9_EtSum7PUTowers','L1Jet3x9_RawEt','L1Jet3x9_EtSum7PUTowers'
params = {'n_estimators': 1000,
'max_depth': 5,
'learning_rate': 0.01}
reg = xgb.XGBRegressor(
random_state=randInt,
**params)
reg.fit(trainData[trainVars], trainData['PFJetEtCorr'])
testPrediction = reg.predict(testData[trainVars])
trainPrediction = reg.predict(trainData[trainVars])
rmse = np.sqrt(MSE(testData['PFJetEtCorr'], testPrediction))
error = ("RMSE=%f")%rmse
print(error)
rank=reg.feature_importances_
feat_name=['L1JetTowerIEtaAbs','L1TJetDefault_PUS','L1JetDefault_PU','L1Jet9x9_RawEt','L1Jet9x9_EtSum7PUTowers','L1Jet7x9_RawEt','L1Jet7x9_EtSum7PUTowers','L1Jet5x9_RawEt','L1Jet5x9_EtSum7PUTowers','L1Jet3x9_RawEt','L1Jet3x9_EtSum7PUTowers']
plt.barh(feat_name,rank)
xlbl=("Feature Ranking for All Jets (All IEtas)")
plt.xlabel(xlbl)
plt.figure().clear()
rankdf=pd.DataFrame(rank)
rankdf.index=['L1JetTowerIEtaAbs','L1TJetDefault_PUS','L1JetDefault_PU','L1Jet9x9_RawEt','L1Jet9x9_EtSum7PUTowers','L1Jet7x9_RawEt','L1Jet7x9_EtSum7PUTowers','L1Jet5x9_RawEt','L1Jet5x9_EtSum7PUTowers','L1Jet3x9_RawEt','L1Jet3x9_EtSum7PUTowers']
print(rankdf)
meanpr=[]
dpusmean=[]
meantr=[]
dpustmean=[]
sdev=[]
sdevtr=[]
res=[]
restr=[]
dpussdev=[]
dpustsdev=[]
dpusres=[]
dpustres=[]
meanout=[]
sdevout=[]
resout=[]
meantrout=[]
dpusmeanout=[]
dpustmeanout=[]
sdevtrout=[]
dpussdev=[]
dpustsdev=[]
dpussdevout=[]
dpustsdevout=[]
restrout=[]
dpusresout=[]
dpustresout=[]
rmsetot=[]
checker = True
testData = testData.reset_index()
del testData['index']
trainData = trainData.reset_index()
del trainData['index']
pred=pd.DataFrame(testPrediction)
pred.columns=['pred']
predT=pd.DataFrame(trainPrediction)
predT.columns=['predT']
frames=[testData,pred]
frames1=[trainData,predT]
pred_all=pd.concat(frames,axis=1)
predT_all=pd.concat(frames1,axis=1)
if checker:
int_pred=pred_all.loc[pred_all['PFJetEtCorr']>=60]
pred_all=int_pred.loc[int_pred['PFJetEtCorr']<=90]
int_predT=predT_all.loc[predT_all['PFJetEtCorr']>=60]
predT_all=int_predT.loc[int_predT['PFJetEtCorr']<=90]
for n in range(1,40):
if n==29:
continue
fin1=pred_all.loc[pred_all['L1JetTowerIEtaAbs']==n]
fin2=predT_all.loc[predT_all['L1JetTowerIEtaAbs']==n]
PFJet=fin1['PFJetEtCorr']
prediction=fin1['pred']
test1=prediction/PFJet
test2=test1.values
pred=pd.DataFrame(test2)
pred.columns=['pred']
ratio_pred=pred.loc[pred['pred']<=3]
test_pred=ratio_pred.values
mean_pred=np.mean(test_pred)
std_pred=np.std(test_pred)
res_pred=std_pred/mean_pred
meanpr.append(mean_pred)
sdev.append(std_pred)
res.append(res_pred)
L1Jet_DPUS=fin1['L1JetDefault_EtPUS']
test3=L1Jet_DPUS/PFJet
test4=test3.values
dpus=pd.DataFrame(test4)
dpus.columns=['dpus']
dpus_ratio=dpus.loc[dpus['dpus']<=3]
dpus_test=dpus_ratio.values
mean_dpus=np.mean(dpus_test)
std_dpus=np.std(dpus_test)
res_dpus=std_dpus/mean_dpus
dpusmean.append(mean_dpus)
dpussdev.append(std_dpus)
dpusres.append(res_dpus)
PFJet_Training=fin2['PFJetEtCorr']
L1Jet_DPUS_Training=fin2['L1JetDefault_EtPUS']
train1=L1Jet_DPUS_Training/PFJet_Training
train2=train1.values
dpust=pd.DataFrame(train2)
dpust.columns=['dpust']
dpust_ratio=dpust.loc[dpust['dpust']<=3]
dpus_train=dpust_ratio.values
mean_dpust=np.mean(dpus_train)
std_dpust=np.std(dpus_train)
res_dpust=std_dpust/mean_dpust
dpustmean.append(mean_dpust)
dpustsdev.append(std_dpust)
dpustres.append(res_dpust)
predictionT=fin2['predT']
train3=predictionT/PFJet_Training
train4=train3.values
pred_training=pd.DataFrame(train4)
pred_training.columns=['pred_training']
ratio_pred_training=pred_training.loc[pred_training['pred_training']<=3]
outlier_training=pred_training.loc[pred_training['pred_training']>=3]
train_pred=ratio_pred_training.values
mean_train_pred=np.mean(train_pred)
std_train_pred=np.std(train_pred)
res_train_pred=std_train_pred/mean_train_pred
meantr.append(mean_train_pred)
sdevtr.append(std_train_pred)
restr.append(res_train_pred)
mylist = list(range(1,40))
x = 28
y=mylist[:x] + mylist[x+1:]
plt.plot(y,meanpr,label='Prediction from BDT(Test)',color='dodgerblue')
plt.plot(y,meantr,label='Prediction from BDT(Training)',color='dodgerblue',ls='--')
plt.plot(y,dpusmean,label='L1TJetDefault_PUS(Test)',color='darkorange')
plt.plot(y,dpustmean,label='L1TJetDefault_PUS(Training)',color='darkorange',ls='--')
# plt.ylim([0.80,1.20])
plt.xlabel("IEta (60<PF Jet pT<90)")
plt.ylabel("Mean = L1T Jet pT/PF Jet pT")
plt.legend()
plt.savefig("Mean vs IEta.png")
plt.figure().clear()
plt.plot(y,res,label='Prediction from BDT(Test)',color='dodgerblue')
plt.plot(y,dpusres,label='L1TJetDefault_PUS(Test)',color='darkorange')
plt.plot(y,restr,label='Prediction from BDT(Training)',color='dodgerblue',ls='--')
plt.plot(y,dpustres,label='L1TJetDefault_PUS(Training)',color='darkorange',ls='--')
# plt.ylim([0,0.7])
plt.xlabel("IEta (60<PF Jet pT<90)")
plt.ylabel("Width/Mean (Width=Sqrt(Variance))")
plt.legend()
plt.savefig("Resolution vs IEta.png")
plt.figure().clear()