def train_lowerq(features):
train_data = features[features['university'] != 'polyu']
test_data = features[features['university'] == 'polyu']
for key, value in university_dict.items():
train_data['university'][train_data['university'] == key] = value
test_data['university'][test_data['university'] == key] = value
lowerq = np.array(list(train_data['lowerq'].values))
test_average = np.array(list(test_data['average'].values))
train_data = train_data.drop(columns=['lowerq', 'upperq'])
test_data = test_data.drop(columns=['lowerq', 'upperq'])
train_data = train_data.get_values()
test_data = test_data.get_values()
regression_model = SBBTree(params=regression_params,
stacking_num=1,
bagging_num=5,
bagging_test_size=0.2,
num_boost_round=10000,
early_stopping_rounds=200)
regression_model.fit(train_data, lowerq)
pred_lowerq = regression_model.predict(test_data)
return get_lowerq(pred_lowerq, test_average), lowerq
params = {
'task': 'train',
'boosting_type': 'gbdt',
'objective': 'regression',
'metric': {'l2', 'auc'},
'num_leaves': 31,
'learning_rate': 0.05,
'feature_fraction': 0.9,
'bagging_fraction': 0.8,
'bagging_freq': 5,
'verbose': 0
}
###############################################################
model = SBBTree(params=params,
stacking_num=5,
bagging_num=3,
bagging_test_size=0.33,
num_boost_round=10000,
early_stopping_rounds=200)
# train 下个月购买次数预测 回归模型
train_features = TrainFeatures.TrainColumns
train_label_BuyNum = 'Label_30_101_BuyNum'
X = TrainFeatures.data_BuyOrNot_FirstTime[train_features].values
y = TrainFeatures.data_BuyOrNot_FirstTime[train_label_BuyNum].values
X_pred = PredFeatures.data_BuyOrNot_FirstTime[train_features].values
model.fit(X, y)
PredFeatures.data_BuyOrNot_FirstTime[train_label_BuyNum] = model.predict(
X_pred)
'num_leaves': 64,
'learning_rate': 0.05,
'feature_fraction': 0.9,
'bagging_fraction': 0.8,
'bagging_freq': 5,
'verbose': 0,
'seed': 2016,
'min_child_weight': 1.5,
'lambda_l2': 10,
'scale_pos_weight': 20
}
###############################################################
model = SBBTree(params=params,
stacking_num=5,
bagging_num=3,
bagging_test_size=0.33,
num_boost_round=10000,
early_stopping_rounds=200,
train_features=train_features)
# train 下个月购买次数预测 回归模型
train_label_BuyNum = 'label_30_101_BuyNums'
train_features = train_features
X = train_data[train_features].values
y = train_data[train_label_BuyNum].values
X_pred = test_data[train_features].values
X_valid = valid_data[train_features].values