def baye_voisin(data, labels, ts):
train_data, test_data, train_labels, test_labels = train_test_split(data, labels, test_size = ts / 100, random_state = 42)
# GAUSSIENNE
g = GaussianBayes()
evaltime()
g.fit(train_data, train_labels)
somme_bayes.append(evaltime())
# - Score:
score_baye = g.score(test_data, test_labels)
Z = g.predict(test_data)
cfmat_bayes = confusion_matrix(test_labels, Z, labels=np.unique(test_labels))
# K-NN
n_neighbors = 10
clf = neighbors.KNeighborsClassifier(n_neighbors, weights='uniform')
evaltime()
clf.fit(train_data, train_labels)
somme_knn.append(evaltime())
# - Score:
score_voisin = clf.score(test_data, test_labels)
Z = clf.predict(test_data)
cfmat_knn = confusion_matrix(test_labels, Z, labels=np.unique(test_labels))
return score_baye, cfmat_bayes, score_voisin, cfmat_knn
def cross_validation(path):
kf = KFold(n_splits=5)
z = np.zeros(100)
data, labels = load_dataset(join(DATA, path[0]))
data = np.array(data)
labels = np.array(labels)
for train, test in kf.split(data):
"""
train_data = []
train_labels = []
test_data = []
test_labels = []
for indice in train:
train_data.append(data[indice])
train_labels.append(labels[indice])
for indice in test:
test_data.append(data[indice])
test_labels.append(labels[indice])
print(train)
print(test)
train_data = np.array(train_data)
train_labels = np.array(train_labels)
test_data = np.array(test_data)
test_labels = np.array(test_labels)"""
#train_data = data[0:800]
#train_labels = labels[0:800]
#test_data = data[801:999]
#test_labels = labels[801:999]
train_data = data[train]
train_labels = labels[train]
test_data = data[test]
test_labels = labels[test]
print(len(train_data))
print(len(train_labels))
print(len(test_data))
print(len(test_labels))
#print(train_data)
#print(train_labels)
#print(test_data)
#print(test_labels)
# GAUSSIENNE
g = GaussianBayes()
g.fit(train_data, train_labels)
# - Score:
score_baye = g.score(test_data, test_labels)
# K-NN
n_neighbors = 10
clf = neighbors.KNeighborsClassifier(n_neighbors, weights='uniform')
clf.fit(train_data, train_labels)
# - Score:
score_voisin = clf.score(test_data, test_labels)
print(score_voisin)
break
def baye_voisin(data, labels, ts):
train_data, test_data, train_labels, test_labels = train_test_split(
data, labels, train_size=ts / 100, random_state=42)
print("test taille:", len(test_data), " train test:", len(train_data),
" test size: ", ts)
# GAUSSIENNE
g = GaussianBayes()
g.fit(train_data, train_labels)
# - Score:
score_baye = g.score(test_data, test_labels)
Z = g.predict(test_data)
cfmat_bayes = confusion_matrix(test_labels,
Z,
labels=np.unique(test_labels))
# K-NN
n_neighbors = 10
clf = neighbors.KNeighborsClassifier(n_neighbors, weights='uniform')
clf.fit(train_data, train_labels)
# - Score:
score_voisin = clf.score(test_data, test_labels)
Z = clf.predict(test_data)
cfmat_knn = confusion_matrix(test_labels, Z, labels=np.unique(test_labels))
return score_baye, cfmat_bayes, score_voisin, cfmat_knn
def main():
train_data, train_labels = load_dataset("./data/train.csv")
test_data, test_labels = load_dataset("./data/test.csv")
#affichage
plot_scatter_hist(train_data, train_labels)
# Instanciation de la classe GaussianB
g = GaussianBayes(priors=[0.3, 0.3, 0.3], diag=True)
# Apprentissage
g.fit(train_data, train_labels)
# Score
score = g.score(test_data, test_labels)
print("precision : {:.2f}".format(score))
input("Press any key to exit...")
(labelstest, total_labels[labelteststart:labeltestend]))
valeurstest = np.concatenate(
(valeurstest, total_data[labelteststart:labeltestend]))
labelstrain = np.concatenate(
(np.concatenate(
(labelstrain, total_labels[labelstrainstart:labelteststart])),
total_labels[labeltestend:labeltrainend]))
valeurstrain = np.concatenate((np.concatenate(
(valeurstrain, total_data[labelstrainstart:labelteststart])),
total_data[labeltestend:labeltrainend]))
g = GaussianBayes(priors=None, diag=False)
# Apprentissage
g.fit(valeurstrain, labelstrain)
# Score
gaussianresult = gaussianresult + g.score(valeurstest, labelstest)
neigh = KNeighborsClassifier(n_neighbors=3,
weights='uniform',
algorithm='brute')
neigh.fit(valeurstrain, labelstrain)
KNNresult = KNNresult + np.sum(
labelstest == neigh.predict(valeurstest)) / len(labelstest)
gaussianresult = gaussianresult / K
KNNresult = KNNresult / K
print("gaussian average precision")
print(gaussianresult)
print("K-NN average precision")
print(KNNresult)
input("Press any key to exit...")
print(i)
labelstest = np.concatenate(
(labelstest, total_labels[labelteststart:labeltestend]))
valeurstest = np.concatenate(
(valeurstest, total_data[labelteststart:labeltestend]))
labelstrain = np.concatenate(
(labelstrain, total_labels[labeltestend:labeltrainend]))
valeurstrain = np.concatenate(
(valeurstrain, total_data[labeltestend:labeltrainend]))
g = GaussianBayes(priors=None, diag=False)
# Apprentissage
g.fit(valeurstrain, labelstrain)
# Score
score = g.score(valeurstest, labelstest)
print("precision : {:.2f}".format(score))
neigh = KNeighborsClassifier(n_neighbors=3,
weights='uniform',
algorithm='brute')
neigh.fit(valeurstrain, labelstrain)
KNeighborsClassifier()
print(np.sum(labelstest == neigh.predict(valeurstest)) / len(labelstest))
#Confusion matrixes
print("confusion matrix for K-NN")
print(confusionmatrixcreator(neigh.predict(valeurstest), labelstest))
print("confusion matrix for the gaussian method")
print(confusionmatrixcreator(g.predict(valeurstest), labelstest))
print(i)
labelstest = np.concatenate(
(labelstest, labeltempfin[0:proportionseparation]))
valeurstest = np.concatenate(
(valeurstest, datatempfin[0:proportionseparation]))
labelstrain = np.concatenate(
(labelstrain, labeltempfin[proportionseparation:100]))
valeurstrain = np.concatenate(
(valeurstrain, datatempfin[proportionseparation:100]))
g = GaussianBayes(priors=None, diag=True)
# Apprentissage
g.fit(valeurstrain, labelstrain)
# Score
score = g.score(valeurstest, labelstest)
print("precision : {:.2f}".format(score))
score = g.score(valeurstrain, labelstrain)
#sur les donnees d'entrainement j'obtiens usuellement des valeurs dans les alentours de 0.73(0.77 pour le cas diagonal)
print("precision : {:.2f}".format(score))
neigh = KNeighborsClassifier(n_neighbors=3,
weights='uniform',
algorithm='brute')
neigh.fit(valeurstrain, labelstrain)
KNeighborsClassifier()
print(np.sum(labelstest == neigh.predict(valeurstest)) / len(labelstest))
#Confusion matrixes
print("confusion matrix for K-NN")
print(confusionmatrixcreator(neigh.predict(valeurstest), labelstest))
#plot_training(color,labels)
# Instanciation de la classe GaussianB
g = GaussianBayes(priors=[0.33, 0.3, 0.326])
# Apprentissage
mu, sig = g.fit(color, labels)
#tab_proba = g.predict(color)
#affichage vraissemblance
#plot_vraissemblance(color,labels,tab_proba)
g.predict(color)
print(g.score(color, labels))
"""
for i in range(n_fleurs):
file_name = "ch"
name = file_name + str(i+1) + ".png"
a, b, c = PretraitementAmeliore(name)
print(a,",",b,",",c,",",0)
for i in range(n_fleurs):
file_name = "oe"
name = file_name + str(i+1) + ".png"
a, b, c = PretraitementAmeliore(name)
print(a,",",b,",",c,",",1)
for i in range(n_fleurs):
file_name = "pe"