使い方忘れるのでメモ．
Scikit-learnのドキュメントのサンプルを少し改変したものとその実行結果．
ソースコード:
grid_search.py
# -*- coding: utf-8 -*-
from sklearn import datasets
from sklearn.cross_validation import train_test_split
from sklearn.grid_search import GridSearchCV
from sklearn.metrics import classification_report, confusion_matrix
from sklearn.svm import SVC
## データの読み込み
digits = datasets.load_digits()
X = digits.data
y = digits.target
## トレーニングデータとテストデータに分割．
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.5, random_state=0)
## チューニングパラメータ
tuned_parameters = [{'kernel': ['rbf'], 'gamma': [1e-3, 1e-4],
'C': [1, 10, 100, 1000]},
{'kernel': ['linear'], 'C': [1, 10, 100, 1000]}]
scores = ['accuracy', 'precision', 'recall']
for score in scores:
print '\n' + '='*50
print score
print '='*50
clf = GridSearchCV(SVC(C=1), tuned_parameters, cv=5, scoring=score, n_jobs=-1)
clf.fit(X_train, y_train)
print "\n+ ベストパラメータ:\n"
print clf.best_estimator_
print"\n+ トレーニングデータでCVした時の平均スコア:\n"
for params, mean_score, all_scores in clf.grid_scores_:
print "{:.3f} (+/- {:.3f}) for {}".format(mean_score, all_scores.std() / 2, params)
print "\n+ テストデータでの識別結果:\n"
y_true, y_pred = y_test, clf.predict(X_test)
print classification_report(y_true, y_pred)
結果:
==================================================
accuracy
==================================================
+ ベストパラメータ:
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0, degree=3, gamma=0.001,
kernel=rbf, max_iter=-1, probability=False, random_state=None,
shrinking=True, tol=0.001, verbose=False)
+ トレーニングデータでCVした時の平均スコア:
0.982 (+/- 0.002) for {'kernel': 'rbf', 'C': 1, 'gamma': 0.001}
0.954 (+/- 0.006) for {'kernel': 'rbf', 'C': 1, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 10, 'gamma': 0.001}
0.981 (+/- 0.003) for {'kernel': 'rbf', 'C': 10, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 100, 'gamma': 0.001}
0.983 (+/- 0.005) for {'kernel': 'rbf', 'C': 100, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 1000, 'gamma': 0.001}
0.983 (+/- 0.005) for {'kernel': 'rbf', 'C': 1000, 'gamma': 0.0001}
0.971 (+/- 0.006) for {'kernel': 'linear', 'C': 1}
0.971 (+/- 0.006) for {'kernel': 'linear', 'C': 10}
0.971 (+/- 0.006) for {'kernel': 'linear', 'C': 100}
0.971 (+/- 0.006) for {'kernel': 'linear', 'C': 1000}
+ テストデータでの識別結果:
precision recall f1-score support
0 1.00 1.00 1.00 89
1 0.97 1.00 0.98 90
2 0.99 0.98 0.98 92
3 1.00 0.99 0.99 93
4 1.00 1.00 1.00 76
5 0.99 0.98 0.99 108
6 0.99 1.00 0.99 89
7 0.99 1.00 0.99 78
8 1.00 0.98 0.99 92
9 0.99 0.99 0.99 92
avg / total 0.99 0.99 0.99 899
==================================================
precision
==================================================
+ ベストパラメータ:
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0, degree=3, gamma=0.001,
kernel=rbf, max_iter=-1, probability=False, random_state=None,
shrinking=True, tol=0.001, verbose=False)
+ トレーニングデータでCVした時の平均スコア:
0.983 (+/- 0.002) for {'kernel': 'rbf', 'C': 1, 'gamma': 0.001}
0.959 (+/- 0.006) for {'kernel': 'rbf', 'C': 1, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 10, 'gamma': 0.001}
0.982 (+/- 0.003) for {'kernel': 'rbf', 'C': 10, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 100, 'gamma': 0.001}
0.985 (+/- 0.004) for {'kernel': 'rbf', 'C': 100, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 1000, 'gamma': 0.001}
0.985 (+/- 0.004) for {'kernel': 'rbf', 'C': 1000, 'gamma': 0.0001}
0.973 (+/- 0.005) for {'kernel': 'linear', 'C': 1}
0.973 (+/- 0.005) for {'kernel': 'linear', 'C': 10}
0.973 (+/- 0.005) for {'kernel': 'linear', 'C': 100}
0.973 (+/- 0.005) for {'kernel': 'linear', 'C': 1000}
+ テストデータでの識別結果:
precision recall f1-score support
0 1.00 1.00 1.00 89
1 0.97 1.00 0.98 90
2 0.99 0.98 0.98 92
3 1.00 0.99 0.99 93
4 1.00 1.00 1.00 76
5 0.99 0.98 0.99 108
6 0.99 1.00 0.99 89
7 0.99 1.00 0.99 78
8 1.00 0.98 0.99 92
9 0.99 0.99 0.99 92
avg / total 0.99 0.99 0.99 899
==================================================
recall
==================================================
+ ベストパラメータ:
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0, degree=3, gamma=0.001,
kernel=rbf, max_iter=-1, probability=False, random_state=None,
shrinking=True, tol=0.001, verbose=False)
+ トレーニングデータでCVした時の平均スコア:
0.982 (+/- 0.002) for {'kernel': 'rbf', 'C': 1, 'gamma': 0.001}
0.954 (+/- 0.006) for {'kernel': 'rbf', 'C': 1, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 10, 'gamma': 0.001}
0.981 (+/- 0.003) for {'kernel': 'rbf', 'C': 10, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 100, 'gamma': 0.001}
0.983 (+/- 0.005) for {'kernel': 'rbf', 'C': 100, 'gamma': 0.0001}
0.986 (+/- 0.002) for {'kernel': 'rbf', 'C': 1000, 'gamma': 0.001}
0.983 (+/- 0.005) for {'kernel': 'rbf', 'C': 1000, 'gamma': 0.0001}
0.971 (+/- 0.006) for {'kernel': 'linear', 'C': 1}
0.971 (+/- 0.006) for {'kernel': 'linear', 'C': 10}
0.971 (+/- 0.006) for {'kernel': 'linear', 'C': 100}
0.971 (+/- 0.006) for {'kernel': 'linear', 'C': 1000}
+ テストデータでの識別結果:
precision recall f1-score support
0 1.00 1.00 1.00 89
1 0.97 1.00 0.98 90
2 0.99 0.98 0.98 92
3 1.00 0.99 0.99 93
4 1.00 1.00 1.00 76
5 0.99 0.98 0.99 108
6 0.99 1.00 0.99 89
7 0.99 1.00 0.99 78
8 1.00 0.98 0.99 92
9 0.99 0.99 0.99 92
avg / total 0.99 0.99 0.99 899
GridSearchCVのパラメータの説明
cv
fold数
scoring
グリードサーチで最適化する値を決められる．
デフォルトでは，
classificationで’accuracy’sklearn.metrics.accuracy_score，
regressionで’r2’sklearn.metrics.r2_scoreが指定されている．
他にも例えばclassificationでは’precision’や’recall’等を指定できる．
詳しくはここ
precision, recall等については朱鷺の杜Wiki
n_jobs
ここに整数値を入れるだけで並列に計算してくれる．
-1でコア数を自動で入れてくれる．