MultiTrain.py

# This script:
#	-Train a BDT from the instruction of the shell file
#	-Compute auc on the 30% test sample and save it in a csv fil for the 3d plot in MultiAnalysis.py
#	-Saves the model for further use


from sys import argv									# To pass variables
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import AdaBoostClassifier
from sklearn.model_selection import train_test_split 	# To use method with same name
from sklearn.metrics import roc_curve, auc 				# Roc: receiver operating characteristic, auc: Area under the ROC Curve
import pandas as pd
import numpy as np
import joblib 											# To save model
import csv 												# Write aoc result in file
from sklearn.neural_network import MLPClassifier		# Import MLPClassifier Neural Network

# Load csv:
pathCSV='/home/mjacquar/TP4b/csv'
X = pd.read_csv(f'{pathCSV}/X.csv')
y = X['sig']


X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, train_size=0.7, random_state=0) # Random state: seed for random assignation of data in the split, done wih kFold
#																																Still present after RFE?
# features = ['B_s0_TAU',			# B lifetime																				no
# 			'MIN_IPCHI2_emu',		# Minimum on the two impact parameters chi2, large if signal (comes from secondary vertex)	yes
# 			'B_s0_IP_OWNPV',		# Impact parameter 																			no, B_s0_IPCHI2_OWNPV instead
# 			'SUM_isolation_emu',	# Isolation: we want nothing in the cone around e or mu 									yes
# 			'B_s0_PT',				# Transverse momentum of B0 																no, B_s0_minPT instead
# 			'LOG1_cosDIRA',			# Direction angle between sum emu and B0 reconstructed -> log(1-cos(DIRA)) 					yes
# 			'B_s0_CDFiso',			# Different measure of isolation, mathematical def in analysis document 					yes
# 			'MAX_PT_emu',			# B0 has high mass -> high p_T for daughter particles 										no
# 			'B_s0_ENDVERTEX_CHI2',	# Quality of the reconstructed decay vertex. 												yes
# 			'DIFF_ETA_emu']			# Absolute difference in pseudorapidity of emu 												no
#																																Added: muminus_ProbNNmu, eplus_ProbNNe, eplus_ETA


features=[  'B_s0_ENDVERTEX_CHI2',
			'B_s0_CDFiso',
			'eplus_ProbNNe',
			'eplus_ETA',
			'B_s0_IPCHI2_OWNPV',
			'B_s0_minPT',
			'muminus_ProbNNmuk',
			'MIN_IPCHI2_emu',
			'SUM_isolation_emu',
			'LOG1_cosDIRA']			# List from the recursive feature elimination

# Load variables:

#BDT:
#algoName = argv[1]
#learningRate=float(argv[2])			# Must convert because argv pass the arguments as strings
#maxDepth=int(argv[3])
#nTrees=int(argv[4])
#dt    = DecisionTreeClassifier(max_depth=maxDepth)														# Define the decision tree
#model = AdaBoostClassifier(dt, algorithm=algoName, n_estimators=nTrees, learning_rate=learningRate)		# Define the model using the decision tree

#MLP:
solver='adam' # Good for datastet of >1000
hidden_layer_sizes=[int(element) for element in argv[1].split(',')] 	# Command will be for example: python3 MultiTrain.py 100,100
																		# Then: hidden_layer_sizes = [100,100]
print(f'hidden_layer_sizes: {hidden_layer_sizes}')
model= MLPClassifier(	solver=solver, 
						hidden_layer_sizes=hidden_layer_sizes, 
						random_state=0)



model.fit(X_train[features], y_train)

hiddenLayerText=""
for element in hidden_layer_sizes:
	hiddenLayerText+= str(element)
	hiddenLayerText+='_'

pathModel='/home/mjacquar/TP4b/model/NeuralNetwork/HyperScan'			# https://medium.com/@harsz89/persist-reuse-trained-machine-learning-models-using-joblib-or-pickle-in-python-76f7e4fd707
#joblib.dump(model,f'{pathModel}/bdt_{algoName}_lr{learningRate}_{nTrees}Trees_{maxDepth}Depth.pkl')		# Save trained model for later
joblib.dump(model,f'{pathModel}/MLP_{hiddenLayerText}.pkl')
#Model saved

y_pred=model.predict_proba(X_test[features])[:,1]
#print("Predictions done")
fpr, tpr, threshold = roc_curve(y_test, y_pred) 	# Use built in fct to compute:  false/true positive read, using the answer and predictions of the test sample
auc = auc(fpr, tpr) 								# Use built in fct to compute area under curve
#print(f'Auc={auc}')
with open(f'{pathCSV}/HyperScanMLP.csv', 'a', newline='') as csvfile: # 'a': append mode to not overwrite
	spamwriter = csv.writer(csvfile, delimiter=' ')
	spamwriter.writerow([features,hidden_layer_sizes,auc])