def sgdClass(X_train, X_test, y_train, y_test, kaggle):
clf = SGDClassifier(loss='hinge',
max_iter=100000,
tol=1e-7,
shuffle=True,
alpha=10,
n_jobs=3)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
mp.finalize(y_test, y_pred, kaggle)
def logRes(X_train, X_test, y_train, y_test, kaggle):
clf = LogisticRegression(C=17,
max_iter=100000,
tol=1e-6,
solver='newton-cg',
multi_class='multinomial',
fit_intercept=False,
random_state=42)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
mp.finalize(y_test, y_pred, kaggle)
def randFost(X_train, X_test, y_train, y_test, kaggle):
clf = RandomForestClassifier(n_estimators=500,
criterion="gini",
bootstrap=True,
min_samples_split=4,
max_depth=55,
max_features='sqrt',
random_state=42)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
mp.finalize(y_test, y_pred, kaggle)
def supportVM(X_train, X_test, y_train, y_test, kaggle):
clf = svm.SVC(kernel='rbf', gamma='scale', C=17)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
mp.finalize(y_test, y_pred, kaggle)
'''
Gave 2nd place private score 0.61691 with public score 0.66017, still overfiting
'''
model.add(Dense(224, activation='relu', input_dim=258))
model.add(Dropout(0.6))
model.add(Dense(128, activation='relu', kernel_regularizer=regularizers.l2(0.001)))
model.add(Dense(64, activation='relu', kernel_regularizer=regularizers.l2(0.001)))
model.add(Dense(8, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer=rmsp, metrics=[mp.f1, 'accuracy'])
model.fit(X_train, y_train, epochs=100, batch_size=256, verbose=2, class_weight=counts, callbacks=[tensorboard])
model.save('3130230_model_' + str(time.time()) + '.h5')
y_pred = model.predict_classes(X_test)
mp.finalize(y_test, y_pred, kaggle)
'''
Gave highest personal public score (0,66467)-> (high overfit) private score 0.61050
model.add(Dense(224, activation='relu', input_dim=258))
model.add(Dropout(0.7))
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.6))
model.add(Dense(8, activation='softmax'))
model.summary()
model.compile(loss='categorical_crossentropy', optimizer=rmsp, metrics=['accuracy'])
model.fit(X_train, y_train, epochs=100, batch_size=256, class_weight=counts, verbose=2)
'''