格點搜尋與隨機搜尋複習

使用 XGBoost 的極端梯度提升

Sergey Fogelson

Head of Data Science, TelevisaUnivision

格點搜尋：複習

針對給定超參數集合做全面搜尋，每組超參數各跑一次
模型數量＝各超參數不同取值數相乘
選擇使交叉驗證評估指標最佳的最終超參數

格點搜尋：範例

import pandas as pd
import xgboost as xgb
import numpy as np
from sklearn.model_selection import GridSearchCV

housing_data = pd.read_csv("ames_housing_trimmed_processed.csv")
X, y = housing_data[housing_data.columns.tolist()[:-1]], 
       housing_data[housing_data.columns.tolist()[-1]
housing_dmatrix = xgb.DMatrix(data=X,label=y)

gbm_param_grid = {'learning_rate': [0.01,0.1,0.5,0.9],
                  'n_estimators': [200],
                  'subsample': [0.3, 0.5, 0.9]}

gbm = xgb.XGBRegressor()
grid_mse = GridSearchCV(estimator=gbm,param_grid=gbm_param_grid, 
            scoring='neg_mean_squared_error', cv=4, verbose=1)
grid_mse.fit(X, y)

print("Best parameters found: ",grid_mse.best_params_)
print("Lowest RMSE found: ", np.sqrt(np.abs(grid_mse.best_score_)))

Best parameters found: {'learning_rate': 0.1, 
'n_estimators': 200, 'subsample': 0.5}
Lowest RMSE found:  28530.1829341

隨機搜尋：複習

為每個欲搜尋的超參數建立（可能無限）取值範圍
設定隨機搜尋要執行的迭代次數
每次迭代時，於各超參數的範圍內隨機抽一組取值，並以該組超參數訓練／評估模型
達到最大迭代次數後，選出評分最佳的超參數組合

隨機搜尋：範例

import pandas as pd
import xgboost as xgb
import numpy as np
from sklearn.model_selection import RandomizedSearchCV
housing_data = pd.read_csv("ames_housing_trimmed_processed.csv")
X,y = housing_data[housing_data.columns.tolist()[:-1]],
      housing_data[housing_data.columns.tolist()[-1]]
housing_dmatrix = xgb.DMatrix(data=X,label=y)

gbm_param_grid = {'learning_rate': np.arange(0.05,1.05,.05),
                  'n_estimators': [200],
                  'subsample': np.arange(0.05,1.05,.05)}

gbm = xgb.XGBRegressor()
randomized_mse = RandomizedSearchCV(estimator=gbm, param_distributions=gbm_param_grid, 
                        n_iter=25, scoring='neg_mean_squared_error', cv=4, verbose=1)
randomized_mse.fit(X, y)

print("Best parameters found: ",randomized_mse.best_params_)
print("Lowest RMSE found: ", np.sqrt(np.abs(randomized_mse.best_score_)))

Best parameters found: {'subsample': 0.60000000000000009,
'n_estimators': 200, 'learning_rate': 0.20000000000000001}
Lowest RMSE found: 28300.2374291

一起來練習吧！

使用 XGBoost 的極端梯度提升