if __name__ == "__main__":
dataset = Dataset(drop_weight=True)
higgs_df = dataset.get_df()
Xtrain, Xtest, ytrain, ytest = dataset.split_train_test(fill_nan=True)
weight_train, weight_test = dataset.get_weight_train_test()
eval_ = Evaluation()
#### Tests de nos différents modèles de bases ####
#random
ypred = pure_random_model(Xtest)
eval_.affichage_score("Random", ytest, ypred, weights=weight_test)
#sratified
stratified = Stratified()
stratified.fit(Xtrain, ytrain)
ypred = stratified.predict(Xtest, ytest)
eval_.affichage_score("Stratified", ytest, ypred, weights=weight_test)
#frequency
frequency = Frequency()
frequency.fit(Xtrain, ytrain)
ypred = frequency.predict(Xtest, ytest)
eval_.affichage_score("Frequency", ytest, ypred, weights=weight_test)
"""
# Classifieur faible : Arbre de décision de profondeur 1 Weight = True
print(" ------------ ENTRAINEMENT RESEAU DE NEURONES ---------------")
for ep in range(n_epoch):
print("EPOCHS : ",ep)
for i, (x, y) in enumerate(train_loader):
model.train()
y = y#.float()
x = x#.double()
pred = model(x)#.double()
print(pred)
loss = criterion(pred, y)
# print('loss', loss ," Prédiction : ", pred, "y True : ",y)
writer.add_scalar('Loss/train', loss, ep)
loss.backward()
optimizer.step()
optimizer.zero_grad()
for i,(x,y) in enumerate(test_loader):
with torch.no_grad():
model.eval()
pred = model(x)
loss = criterion(pred,y)
# print('loss', loss ," Prédiction : ", pred, "y True : ",y)
writer.add_scalar('Loss/validation', loss, ep)
ypred = dataset_train.get_prediction(ynn)
eval_.affichage_score("Neural Network ",ytest,ypred,weights=weight_test)
import random,string,math,csv
import numpy as np
import pandas as pd
import sys
sys.path.insert(0, '/Users/ykarmim/Documents/Cours/Master/M2/REDS/src') # Mettre le path vers src sinon l'import fail.
from dataset import Dataset
from evaluation import Evaluation
from xgboost import XGBClassifier
if __name__ == "__main__":
dataset = Dataset(drop_weight=True)
higgs_df = dataset.get_df()
Xtrain,Xtest, ytrain,ytest = dataset.split_train_test(fill_nan=False) # XgBoost prend en compte les valeurs NaN.
weight_train,weight_test = dataset.get_weight_train_test()
eval_ = Evaluation()
model = XGBClassifier()
model.fit(Xtrain, ytrain)
ypred = model.predict(Xtest)
eval_.affichage_score("XGBoost",ytest,ypred,weights=weight_test)
if __name__ == "__main__":
dataset = Dataset(drop_weight=True)
higgs_df = dataset.get_df()
Xtrain,Xtest, ytrain,ytest = dataset.split_train_test(fill_nan=True)
weight_train,weight_test = dataset.get_weight_train_test()
eval_ = Evaluation()
#Decision tree automatic depth
tree_1 = DecisionTreeClassifier(random_state=0)
tree_1.fit(Xtrain,ytrain)
ypred = tree_1.predict(Xtest)
ypred_train = tree_1.predict(Xtrain)
eval_.affichage_score("Decision Tree automatic depth Weight = False, Test",ytest,ypred,weights=weight_test)
eval_.affichage_score("Decision Tree automatic depth Weight = False, Train",ytrain,ypred_train,weights=weight_train)
# Decision tree fixed depth 5
tree_1 = DecisionTreeClassifier(random_state=0,max_depth=5)
tree_1.fit(Xtrain,ytrain)
ypred = tree_1.predict(Xtest)
ypred_train = tree_1.predict(Xtrain)
eval_.affichage_score("Decision Tree depth 5 Weight = False, Test",ytest,ypred,weights=weight_test)
eval_.affichage_score("Decision Tree depth 5 Weight = False, Train",ytrain,ypred_train,weights=weight_train)
# à faire : cross val pour trouver le meilleur hyper paramètre pour la profondeur maximal de l'arbre
ypred_train = tree_1.predict(Xtrain)
eval_.affichage_score("Decision Tree depth 5 Weight = False, Test",ytest,ypred,weights=weight_test)
eval_.affichage_score("Decision Tree depth 5 Weight = False, Train",ytrain,ypred_train,weights=weight_train)
# à faire : cross val pour trouver le meilleur hyper paramètre pour la profondeur maximal de l'arbre
"""
# Random Forest Bagging bootstrap = True , n_estimator = 100
optim_param = pd.read_csv('results/optimal_parameters_rf.csv')
params = optim_param['params']
params = ast.literal_eval(params[0])
rf = RandomForestClassifier(**params)
rf.fit(Xtrain, ytrain)
ypred = rf.predict(scale(Xtest))
eval_.affichage_score("Random Forest bootstrap = True ",
ytest,
ypred,
weights=weight_test)
"""
rf = RandomForestClassifier(n_estimators=100,bootstrap=False,max_depth=5)
rf.fit(Xtrain,ytrain)
ypred = rf.predict(Xtest)
eval_.affichage_score("Random Forest bootstrap = False ",ytest,ypred,weights=weight_test)
# à faire cross val pour les hyper-paramètres
"""
# Extra Trees
"""Extra trees classifier.
This classifier is very similar to a random forest; in extremely
randomized trees, randomness goes one step further in the way splits are
computed. As in random forests, a random subset of candidate features is
used, but instead of looking for the most discriminative thresholds,
thresholds are drawn at random for each candidate feature and the best of
