xgboostのパラメータをグリッドサーチで決定する

モデルの詳細

Python API Reference — xgboost 0.6 documentation　にあるパラメータのうち、特に影響が大きい物をグリッドサーチで決定します。xgboostの本論文はKDD2016の以下を参照。最近のコンペでは協力なベースラインとしてみんな使ってますね。

www.kdd.org

Tree boosting is a highly effective and widely used machine learning method. In this paper, we describe a scalable end-to-end tree boosting system called XGBoost, which is used widely by data scientists to achieve state-of-the-art results on many machine learning challenges.

コード

from xgboost import XGBRegressor
from sklearn.pipeline import Pipeline
from sklearn.decomposition import PCA, NMF
from sklearn.preprocessing import PolynomialFeatures
from sklearn.feature_selection import SelectKBest, chi2
from sklearn.model_selection import GridSearchCV

pipeline = Pipeline([
    ('clf', XGBRegressor()),
])

params = dict(clf__n_estimators=(10, 20, 30),
              clf__learning_rate=(.1, .2, .3), 
              clf__max_depth=(2, 3, 4, 5),
              clf__min_child_weight=(.5, .75, 1.0))

grid_search = GridSearchCV(pipeline, param_grid=params).fit(train_X, train_y)
predictions = grid_search.predict(test_X)

参考スライド

XGBoost: the algorithm that wins every competition from Jaroslaw Szymczak

www.slideshare.net