def clsf2_by_clst(Xpart_cf, Xpart_ct):
"""
Clustering is performed and then, classification performed by clustered indices.
"""
cl_model = cluster.KMeans(n_clusters=nb_classes)
cl_model.fit(Xpart_ct)
yint = cl_model.predict(Xpart_ct)
X_train, X_test, y_train, y_test = \
model_selection.train_test_split( Xpart_cf, yint, test_size = 0.2)
model = tree.DecisionTreeClassifier()
model.fit(X_train, y_train)
dt_score = model.score(X_test, y_test)
print("DT-C:", dt_score)
model = svm.SVC(kernel='linear')
model.fit(X_train, y_train)
sv_score = model.score(X_test, y_test)
print("SVC:", sv_score)
model = kkeras.MLPC([Xpart_cf.shape[1], 30, 10, nb_classes])
model.fit(X_train, y_train, X_test, y_test, nb_classes)
mlp_score = model.score(X_test, y_test)
print("MLP:", mlp_score)
return dt_score, sv_score, mlp_score
def clst(X_train, y_train, X_test, y_test, nb_classes):
model = tree.DecisionTreeClassifier()
model.fit(X_train, y_train)
dt_score = model.score(X_test, y_test)
print("DT-C:", dt_score)
model = svm.SVC(kernel='linear')
model.fit(X_train, y_train)
sv_score = model.score(X_test, y_test)
print("SVC:", sv_score)
model = kkeras.MLPC([X_train.shape[1], 30, 10, nb_classes])
model.fit(X_train, y_train, X_test, y_test, nb_classes)
mlp_score = model.score(X_test, y_test)
print("DNN:", mlp_score)
model = kkeras.CNNC(param_d["n_cv_flt"],
param_d["n_cv_ln"],
param_d["cv_activation"],
l=[X_train.shape[1], 30, 10, nb_classes])
model.fit(X_train, y_train, X_test, y_test, nb_classes)
cnn_score = model.score(X_test, y_test)
print("DCNN:", cnn_score)
model = ensemble.RandomForestClassifier(n_estimators=10)
model.fit(X_train, y_train)
rf_score = model.score(X_test, y_test)
print("RF:", rf_score)
return dt_score, sv_score, mlp_score, cnn_score, rf_score
def _clst_r0(X_train, y_train, X_test, y_test, nb_classes, disp=True):
model = tree.DecisionTreeClassifier()
model.fit(X_train, y_train)
dt_score = model.score(X_test, y_test)
model = svm.SVC(C=param_d['SVC:C'], gamma=param_d['SVC:gamma'])
# print(model)
model.fit(X_train, y_train)
sv_score = model.score(X_test, y_test)
model = kkeras.MLPC(
[X_train.shape[1], param_d["DNN:H1"], param_d["DNN:H2"], nb_classes])
model.fit(X_train, y_train, X_test, y_test, nb_classes)
mlp_score = model.score(X_test, y_test)
model = kkeras.CNNC(
param_d["n_cv_flt"],
param_d["n_cv_ln"],
param_d["cv_activation"],
l=[X_train.shape[1], param_d["DNN:H1"], param_d["DNN:H2"], nb_classes])
model.fit(X_train, y_train, X_test, y_test, nb_classes)
cnn_score = model.score(X_test, y_test)
model = ensemble.RandomForestClassifier(
n_estimators=param_d['RF:n_estimators'],
oob_score=param_d['RF:oob_score'])
# print(model)
model.fit(X_train, y_train)
rf_score = model.score(X_test, y_test)
if disp:
print("DT-C:", dt_score)
print("SVC:", sv_score)
print("DNN:", mlp_score)
print("DCNN:", cnn_score)
print("RF:", rf_score)
return dt_score, sv_score, mlp_score, cnn_score, rf_score
def clsf2_by_yint(X1part, yint):
"""
classification is performed by yint
"""
X_train, X_test, y_train, y_test = \
model_selection.train_test_split( X1part, yint, test_size = 0.2)
model = tree.DecisionTreeClassifier()
model.fit(X_train, y_train)
dt_score = model.score(X_test, y_test)
print("DT:", dt_score)
model = svm.SVC(kernel='linear')
model.fit(X_train, y_train)
sv_score = model.score(X_test, y_test)
print("SVC:", sv_score)
model = kkeras.MLPC([X1part.shape[1], 30, 10, nb_classes])
model.fit(X_train, y_train, X_test, y_test, nb_classes)
mlp_score = model.score(X_test, y_test)
print("MLP:", mlp_score)
return dt_score, sv_score, mlp_score
def clst(X_train,
y_train,
X_test,
y_test,
nb_classes,
disp=True,
confusion_matric_return=False,
matthews_corrcoef_return=False):
"""
dt_score, sv_score, mlp_score, cnn_score, rf_score =
clst(X_train, y_train, X_test, y_test, nb_classes, disp=True,
confusion_matric_return=False):
dt_score, sv_score, mlp_score, cnn_score, rf_score,
dt_cf, sv_cf, mlp_cf, cnn_cf, rf_cf =
clst(X_train, y_train, X_test, y_test, nb_classes, disp=True,
confusion_matric_return=True):
Claustering by multiple classification methods
Inputs
======
confusion_matric_return, True or False
Whether it will return a confusion matric or not.
"""
model = tree.DecisionTreeClassifier()
model.fit(X_train, y_train)
dt_score = model.score(X_test, y_test)
y_pred = model.predict(X_test)
dt_cf = metrics.confusion_matrix(y_test, y_pred)
if matthews_corrcoef_return:
dt_mc = metrics.matthews_corrcoef(y_test, y_pred)
model = svm.SVC(C=param_d['SVC:C'], gamma=param_d['SVC:gamma'])
# print(model)
model.fit(X_train, y_train)
sv_score = model.score(X_test, y_test)
y_pred = model.predict(X_test)
sv_cf = metrics.confusion_matrix(y_test, y_pred)
if matthews_corrcoef_return:
sv_mc = metrics.matthews_corrcoef(y_test, y_pred)
model = kkeras.MLPC(
[X_train.shape[1], param_d["DNN:H1"], param_d["DNN:H2"], nb_classes])
model.fit(X_train, y_train, X_test, y_test, nb_classes)
mlp_score = model.score(X_test, y_test)
y_pred = model.predict(X_test)
mlp_cf = metrics.confusion_matrix(y_test, y_pred)
if matthews_corrcoef_return:
mlp_mc = metrics.matthews_corrcoef(y_test, y_pred)
model = kkeras.CNNC(
param_d["n_cv_flt"],
param_d["n_cv_ln"],
param_d["cv_activation"],
l=[X_train.shape[1], param_d["DNN:H1"], param_d["DNN:H2"], nb_classes])
model.fit(X_train, y_train, X_test, y_test, nb_classes)
cnn_score = model.score(X_test, y_test)
y_pred = model.predict(X_test)
cnn_cf = metrics.confusion_matrix(y_test, y_pred)
if matthews_corrcoef_return:
cnn_mc = metrics.matthews_corrcoef(y_test, y_pred)
model = ensemble.RandomForestClassifier(
n_estimators=param_d['RF:n_estimators'],
oob_score=param_d['RF:oob_score'])
# print(model)
model.fit(X_train, y_train)
rf_score = model.score(X_test, y_test)
y_pred = model.predict(X_test)
rf_cf = metrics.confusion_matrix(y_test, y_pred)
if matthews_corrcoef_return:
rf_mc = metrics.matthews_corrcoef(y_test, y_pred)
if disp:
print('Acrracuy, matthews corrcoef, confusion metrics')
print("DT-C:", dt_score, dt_mc)
print(dt_cf)
print("SVC:", sv_score, sv_mc)
print(sv_cf)
print("DNN:", mlp_score, mlp_mc)
print(mlp_cf)
print("DCNN:", cnn_score, cnn_mc)
print(cnn_cf)
print("RF:", rf_score, rf_mc)
print(rf_cf)
return_l = [dt_score, sv_score, mlp_score, cnn_score, rf_score]
if confusion_matric_return:
return_l.extend([dt_cf, sv_cf, mlp_cf, cnn_cf, rf_cf])
if matthews_corrcoef_return:
return_l.extend([dt_mc, sv_mc, mlp_mc, cnn_mc, rf_mc])
return return_l