def adaptive_test(X_train,
X_test,
y_train,
y_test,
n_iter=100,
n_jobs=4,
cv=3):
classifier = tree.DecisionTreeClassifier()
auc = make_scorer(roc_auc_score)
rand_list = {
"max_leaf_nodes": [2, 3, 4, 5],
"max_features": stats.uniform(0.5, 0.25)
}
rand_search = RandomizedSearchCV(classifier,
param_distributions=rand_list,
n_iter=n_iter,
n_jobs=n_jobs,
cv=cv,
random_state=2017,
scoring=auc)
rand_search.fit(X_train, y_train)
y_pre = rand_search.predict(X_test)
return rand_search_result(rand_search.best_params_, y_test, y_pre)
def adaptive_test(X_train,
X_test,
y_train,
y_test,
n_iter=100,
n_jobs=4,
cv=3):
classifier = neighbors.KNeighborsClassifier()
auc = make_scorer(roc_auc_score)
rand_list = {
"n_neighbors": [4, 5, 6, 7, 8, 9],
"leaf_size": [25, 30, 35],
"algorithm": ['ball_tree', 'kd_tree', 'brute'],
"p": [1, 2]
}
rand_search = RandomizedSearchCV(classifier,
param_distributions=rand_list,
n_iter=n_iter,
n_jobs=n_jobs,
cv=cv,
random_state=2017,
scoring=auc)
rand_search.fit(X_train, y_train)
y_pre = rand_search.predict(X_test)
return rand_search_result(rand_search.best_params_, y_test, y_pre)
def adaptive_test(X_train,
X_test,
y_train,
y_test,
n_iter=100,
n_jobs=4,
cv=3):
cat_features = [0, 1, 6]
classifier = CatBoost()
auc = make_scorer(roc_auc_score)
rand_list = {
'iterations': [1, 2, 5, 10, 20, 50, 100, 200],
'depth': [*range(1, 11)],
'learning_rate':
[.005, .01, .02, .05, .1, .2, .5, 1.0, 2.0, 5.0, 10.0],
'loss_function': ['Logloss']
}
rand_search = RandomizedSearchCV(classifier,
param_distributions=rand_list,
n_iter=n_iter,
n_jobs=n_jobs,
cv=cv,
random_state=2017,
scoring=auc)
rand_search.fit(X_train, y_train)
y_pre = rand_search.best_estimator_.predict(X_test,
prediction_type='Class')
return rand_search_result(rand_search.best_params_, y_test, y_pre)
def adaptive_test(X_train,
X_test,
y_train,
y_test,
n_iter=100,
n_jobs=4,
cv=3):
classifier = naive_bayes.BernoulliNB()
auc = make_scorer(roc_auc_score)
rand_list = {
"alpha": stats.uniform(0.5, 1),
"binarize": stats.uniform(0.0, 0.5)
}
rand_search = RandomizedSearchCV(classifier,
param_distributions=rand_list,
n_iter=n_iter,
n_jobs=n_jobs,
cv=cv,
random_state=2017,
scoring=auc)
rand_search.fit(X_train, y_train)
y_pre = rand_search.predict(X_test)
return rand_search_result(rand_search.best_params_, y_test, y_pre)
def adaptive_test(X_train,
X_test,
y_train,
y_test,
n_iter=100,
n_jobs=4,
cv=3):
classifier = svm.SVC()
auc = make_scorer(roc_auc_score)
rand_list = {"C": stats.uniform(0.5, 10), "gamma": stats.uniform(0.05, 1)}
rand_search = RandomizedSearchCV(classifier,
param_distributions=rand_list,
n_iter=n_iter,
n_jobs=n_jobs,
cv=cv,
random_state=2017,
scoring=auc)
rand_search.fit(X_train, y_train)
y_pre = rand_search.predict(X_test)
return rand_search_result(rand_search.best_params_, y_test, y_pre)
def adaptive_test(X_train,
X_test,
y_train,
y_test,
n_iter=100,
n_jobs=4,
cv=3):
classifier = linear_model.LogisticRegression()
auc = make_scorer(roc_auc_score)
rand_list = {
"tol": stats.uniform(1e-6, 1e-4),
"C": stats.uniform(0.5, 1),
"max_iter": [100, 150, 200]
}
rand_search = RandomizedSearchCV(classifier,
param_distributions=rand_list,
n_iter=n_iter,
n_jobs=n_jobs,
cv=cv,
random_state=2017,
scoring=auc)
rand_search.fit(X_train, y_train)
y_pre = rand_search.predict(X_test)
return rand_search_result(rand_search.best_params_, y_test, y_pre)
def adaptive_test(X_train, X_test, y_train, y_test, n_iter=100, n_jobs=4, cv=3):
classifier = XGBClassifier()
auc = make_scorer(roc_auc_score)
rand_list = {"learning_rate": stats.uniform(1e-6, 1e-2),
"n_estimators" : [100, 300, 500]}
rand_search = RandomizedSearchCV(classifier, param_distributions=rand_list,
n_iter=n_iter, n_jobs=n_jobs, cv=cv, random_state=2017, scoring=auc)
rand_search.fit(X_train, y_train)
y_pre = rand_search.predict(X_test)
return rand_search_result(rand_search.best_params_, y_test, y_pre)
def adaptive_test(X_train, X_test, y_train, y_test, n_iter=50, n_jobs=4, cv=2):
classifier = SklearnHelper()
auc = make_scorer(roc_auc_score)
rand_list = {"v": [1], "threshold": stats.uniform(-.99, 1.99)}
rand_search = RandomizedSearchCV(classifier,
param_distributions=rand_list,
n_iter=n_iter,
n_jobs=n_jobs,
cv=cv,
random_state=42,
scoring=auc)
rand_search.fit(X_train, np.int8(y_train))
y_pre = rand_search.predict(X_test)
return rand_search_result(rand_search.best_params_, y_test, y_pre)
def basic_test():
classifier = fris_stolp(.5, 1)
classifier.fit(X_train.values, np.int8(y_train.values))
res = classifier.predict(X_test.values)
print(res)
return rand_search_result({}, y_test, res)