時系列データでのGrid Search

はじめに

意外と情報が見つからなかったのでまとめてみました。
sklearn.model_selection.TimeSeriesSplitを用いた実行例を以下に示します。

実行例

import numpy as np
import xgboost as xgb
from sklearn.model_selection import TimeSeriesSplit, GridSearchCV

X = np.array([[40, 15, 36, 11, 100, 22, 21, 76, 3, 32], 
              [3.1, 3.5, 1.3, 2.1, 8.3, 1.1, 3.4, 2.0, 2.4, 5.6]]).T
y = np.array([3, 21, 13, 7, 4, 1, 19, 2, 1, 1])

tscv = TimeSeriesSplit(n_splits=3)
for train, test in tscv.split(X):
    print(train, test)

[0 1 2 3] [4 5]
[0 1 2 3 4 5] [6 7]
[0 1 2 3 4 5 6 7] [8 9]

XGBoostでの実行例

import numpy as np
import xgboost as xgb
from sklearn.model_selection import TimeSeriesSplit, GridSearchCV

X = np.array([[40, 15, 36, 11, 100, 22, 21, 76, 3, 32], 
              [3.1, 3.5, 1.3, 2.1, 8.3, 1.1, 3.4, 2.0, 2.4, 5.6]]).T
y = np.array([3, 21, 13, 7, 4, 1, 19, 2, 1, 1])

model = xgb.XGBRegressor()
params = {
    "learning_rate"   : [0.1,0.3,0.5],
    "max_depth"       : [2,3,5],
}

tscv = TimeSeriesSplit(n_splits=3).split(X)
gscv = GridSearchCV(estimator=model,
                    cv=tscv,
                    param_grid=params)

GridSearchCV(cv=<generator object TimeSeriesSplit.split at 0x7f2a5ee6c830>,
       error_score='raise',
       estimator=XGBRegressor(base_score=0.5, booster='gbtree', colsample_bylevel=1,
       colsample_bytree=1, gamma=0, learning_rate=0.1, max_delta_step=0,
       max_depth=3, min_child_weight=1, missing=None, n_estimators=100,
       n_jobs=1, nthread=None, objective='reg:linear', random_state=0,
       reg_alpha=0, reg_lambda=1, scale_pos_weight=1, seed=None,
       silent=True, subsample=1),
       fit_params={}, iid=True, n_jobs=1,
       param_grid={'learning_rate': [0.1, 0.3, 0.5], 'max_depth': [2, 3, 5]},
       pre_dispatch='2*n_jobs', refit=True, return_train_score=True,
       scoring=None, verbose=0)

参考