# 학습 데이터셋
train_X = any_football_data.drop(label_columns, axis=1)
train_Y = any_football_data['result_type'] # 경기 결과 (승=0, 무=1, 패=2)
# 특성 처리
feature_pipeline = ColumnTransformer([
("num", StandardScaler(), num_attribs),
("cat", OneHotEncoder(), cat_attribs_with_camp),
])
# 모델 훈련
prepared_train = feature_pipeline.fit_transform(train_X)
print(f'prepared_train data_set shape: {prepared_train.shape}')
start_prepared = time.time()
knn_clf = KNeighborsClassifier(weights='distance', n_neighbors=19)
knn_clf.fit(prepared_train, train_Y)
end_prepared = time.time()
print(f'elapsed time: {end_prepared - start_prepared} sec \n')
full_pipeline = Pipeline([("preparation", feature_pipeline),
("predictor", knn_clf)])
# 저장
dir_name = datetime.today().strftime("knn_%Y%m%d-%H%M%S")
utility.save_sklearn_model(full_pipeline, dir_name, "knn_clf_result_any.pkl")
sgd_home_pred = np.round(sgd_clf_4_5_any.predict_proba(home_train), decimals=3)
sgd_away_pred = np.round(sgd_clf_4_5_any.predict_proba(away_train), decimals=3)
home_prepared_train = np.c_[np.ones(len(home_train)),
np.zeros(len(home_train)), knn_home_pred,
rft_home_pred, sgd_home_pred]
away_prepared_train = np.c_[np.zeros(len(away_train)),
np.ones(len(away_train)), knn_away_pred,
rft_away_pred, sgd_away_pred]
train_X = np.r_[home_prepared_train, away_prepared_train]
train_Y = np.r_[home_result_4_5_over_Y, away_result_4_5_over_Y]
# SGDClassifier 학습
start_prepared = time.time()
knn_clf_4_5 = KNeighborsClassifier(weights='uniform', n_neighbors=100)
knn_clf_4_5.fit(train_X, train_Y)
end_prepared = time.time()
print(
f'4.5 knn predictor elapsed time: {end_prepared - start_prepared} sec \n')
# Save
dir_name = datetime.today().strftime("fin_%Y%m%d-%H%M%S")
utility.save_sklearn_model(knn_clf_1_5, dir_name, "fin_knn_clf_1_5.pkl")
utility.save_sklearn_model(knn_clf_2_5, dir_name, "fin_knn_clf_2_5.pkl")
utility.save_sklearn_model(knn_clf_3_5, dir_name, "fin_knn_clf_3_5.pkl")
utility.save_sklearn_model(knn_clf_4_5, dir_name, "fin_knn_clf_4_5.pkl")
lin_away_pred = np.round(lin_reg_any.predict(away_train), decimals=3)
# prepare data
# league_cat = utility.LeagueTypeCategorize()
# leagues = league_cat.transform(home_train.copy())
home_prepared_train = np.c_[np.ones(len(home_train)),
np.zeros(len(home_train)), knn_home_pred,
lin_home_pred]
away_prepared_train = np.c_[np.zeros(len(away_train)),
np.ones(len(away_train)), knn_away_pred,
lin_away_pred]
train_X = np.r_[home_prepared_train, away_prepared_train]
train_Y = np.r_[home_score_Y, away_score_Y]
# LinearRegression 학습
start_prepared = time.time()
fin_lin_reg = LinearRegression()
fin_lin_reg.fit(train_X, train_Y) # 훈련
end_prepared = time.time()
print(
f'lin score predictor elapsed time: {end_prepared - start_prepared} sec \n'
)
# Save
dir_name = datetime.today().strftime("fin_%Y%m%d-%H%M%S")
utility.save_sklearn_model(fin_lin_reg, dir_name, "fin_lin_reg_score.pkl")
train_X = np.r_[home_prepared_train, away_prepared_train]
train_Y = np.r_[home_result_Y, away_result_Y]
# SGDClassifier 학습
# start_prepared = time.time()
#
# sgd_clf = SGDClassifier(max_iter=3000, tol=1e-3, random_state=42, alpha=0.01, loss='log')
# sgd_clf.fit(train_X, train_Y)
#
# end_prepared = time.time()
#
# print(f'sgd result predictor elapsed time: {end_prepared - start_prepared} sec \n')
# KNeighborsClassifier 학습
start_prepared = time.time()
knn_clf = KNeighborsClassifier(weights='uniform', n_neighbors=100)
knn_clf.fit(train_X, train_Y)
end_prepared = time.time()
print(
f'knn result predictor elapsed time: {end_prepared - start_prepared} sec \n'
)
# Save
dir_name = datetime.today().strftime("fin_%Y%m%d-%H%M%S")
# utility.save_sklearn_model(sgd_clf, dir_name, "fin_sgd_clf_result.pkl")
utility.save_sklearn_model(knn_clf, dir_name, "fin_knn_clf_result.pkl")
# 학습 데이터셋
train_X = any_football_data.drop(label_columns, axis=1)
train_Y = any_football_data['score'] # 경기 결과 (승=0, 무=1, 패=2)
# 특성 처리
feature_pipeline = ColumnTransformer([
("num", StandardScaler(), num_attribs),
("cat", OneHotEncoder(), cat_attribs_with_camp),
])
# 모델 훈련
prepared_train = feature_pipeline.fit_transform(train_X)
print(f'prepared_train data_set shape: {prepared_train.shape}')
start_prepared = time.time()
knn_reg = KNeighborsRegressor(n_neighbors=19, weights="distance")
knn_reg.fit(prepared_train, train_Y) # 훈련
end_prepared = time.time()
print(f'elapsed time: {end_prepared - start_prepared} sec \n')
full_pipeline = Pipeline([("preparation", feature_pipeline),
("predictor", knn_reg)])
# 저장
dir_name = datetime.today().strftime("knn_%Y%m%d-%H%M%S")
utility.save_sklearn_model(full_pipeline, dir_name, "knn_reg_score_any.pkl")
knn_clf_3_5 = KNeighborsClassifier(weights='distance', n_neighbors=19)
knn_clf_3_5.fit(prepared_train, train_3_5_Y)
end_prepared = time.time()
print(f'3.5 predictor elapsed time: {end_prepared - start_prepared} sec \n')
full_pipeline_3_5 = Pipeline([("preparation", feature_pipeline),
("predictor", knn_clf_3_5)])
# 모델 훈련 (4.5)
start_prepared = time.time()
knn_clf_4_5 = KNeighborsClassifier(weights='distance', n_neighbors=19)
knn_clf_4_5.fit(prepared_train, train_4_5_Y)
end_prepared = time.time()
print(f'4.5 predictor elapsed time: {end_prepared - start_prepared} sec \n')
full_pipeline_4_5 = Pipeline([("preparation", feature_pipeline),
("predictor", knn_clf_4_5)])
# 저장
dir_name = datetime.today().strftime("knn_%Y%m%d-%H%M%S")
utility.save_sklearn_model(full_pipeline_1_5, dir_name, "knn_clf_1_5_any.pkl")
utility.save_sklearn_model(full_pipeline_2_5, dir_name, "knn_clf_2_5_any.pkl")
utility.save_sklearn_model(full_pipeline_3_5, dir_name, "knn_clf_3_5_any.pkl")
utility.save_sklearn_model(full_pipeline_4_5, dir_name, "knn_clf_4_5_any.pkl")
("cat", OneHotEncoder(), cat_attribs_with_camp),
])
feature_pipeline = Pipeline([
('columns', columns_pipeline_home),
])
# 모델 훈련
prepared_train = feature_pipeline.fit_transform(train_X)
print(f'prepared_train data_set shape: {prepared_train.shape}')
start_prepared = time.time()
sgd_clf = SGDClassifier(max_iter=2000,
tol=1e-3,
random_state=42,
alpha=0.01,
loss='log')
sgd_clf.fit(prepared_train, train_Y)
end_prepared = time.time()
print(f'elapsed time: {end_prepared - start_prepared} sec \n')
full_pipeline = Pipeline([("preparation", feature_pipeline),
("predictor", sgd_clf)])
# 저장
dir_name = datetime.today().strftime("sgd_%Y%m%d-%H%M%S")
utility.save_sklearn_model(full_pipeline, dir_name, "sgd_clf_bts_any.pkl")