個人用scikit-learn

回帰

線形回帰

from sklearn.linear_model import LinearRegression
lin_reg = LinearRegression()
lin_reg.fit(X, y)
y_pred = lin_reg.predict(X)
lin_reg.score(X, y)  #R2スコア
lin_reg.get_params()
lin_reg.set_params(**params)

Ridge回帰（L2正則化）

from sklearn.linear_model import Ridge
ridge = Ridge(alpha=1)
ridge.fit(X, y)
y_pred = ridge.predict(X)
ridge.score(X, y)

グリッドサーチによるモデル選択

from sklearn.linear_model import Ridge
from sklearn.model_selection import GridSearchCV
param_grid = {'alpha': [0.1, 0.25, 0.5, 0.75, 0.9]}
grid_search = GridSearchCV(Ridge(), param_grid=param_grid, cv=5)
grid_search.fit(X, y)
best_model = grid_search.best_estimator_
print(grid_search.best_params_)
print(grid_search.best_score_)

Lasso回帰（L1正則化）

from sklearn.linear_model import Lasso
lasso = Ridge(alpha=0.1)
lasso.fit(X, y)
y_pred = lasso.predict(X)
lasso.score(X, y)

グリッドサーチによるモデル選択

from sklearn.linear_model import Lasso
from sklearn.model_selection import GridSearchCV
param_grid = {'alpha': [0.1, 0.25, 0.5, 0.75, 0.9]}
grid_search = GridSearchCV(Lasso(), param_grid=param_grid, cv=5)
grid_search.fit(X, y)
best_model = grid_search.best_estimator_
print(grid_search.best_params_)
print(grid_search.best_score_)

ElasticNet

グリッドサーチによるモデル選択

from sklearn.linear_model import ElasticNet
from sklearn.model_selection import GridSearchCV
param_grid = {'alpha': [0.1, 0.5, 0.9], 'l1_ratio': [0.01, 0.1, 0.5, 0.9, 0.99]}
grid_search = GridSearchCV(ElasticNet(), param_grid=param_grid, cv=5)
grid_search.fit(X, y)
best_model = grid_search.best_estimator_
print(grid_search.best_params_)
print(grid_search.best_score_)

決定木

インポート

import matplotlib.pyplot as plt
from sklearn.datasets import load_iris
from sklearn.tree import DecisionTreeClassifier, plot_tree

学習と予測

X = data
y = target
tree_clf = DecisionTreeClassifier(max_depth=2)
tree_clf.fit(X, y)
tree_clf.pridict_proba(new_data)
tree_clf.pridict(new_data)

決定木の可視化

plot_tree(tree_clf, feature_names=iris.feature_names, filled=True)
plt.show()

ランダムフォレスト

rnd_clf = RondomForestClassifier()
rnd_reg = RondomForestRegressor()

ハイパーパラメータ	選択肢	default
n_estimators	int型	10
criterion	gini、entropy	gini
max_depth	int型 or None	Nonemin_samples_split
min_samples_leaf	int、float型	1
min_weight_fraction_leaf	float型	0
max_features	int、float型、None、auto、sqrt、log2	auto
max_leaf_nodes	int型 or None	None
min_impurity_decrease	float型	0
min_impurity_split	float型	1e-7
bootstrap	bool型	True
oob_score	bool型	False
n_jobs	int型 or None	None
random_state	int型、RandomStateinstanceorNone	None
verbose	int型	0
warm_start	bool型	False
class_weight	辞書型、balanced、balanced_subsampleorNone	None

詳しくは
https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html

データ変換

最大最小スケーリング（正規化）

パラメータfeature_rangeは省略することもできる。その場合は0から１になる。

from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler(feature_range=(0, 1))
X_scaled = scaler.fit_transform(X)

標準化

from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

モデル評価、選択

訓練データと検証データに分割

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

K分割交差検証

from sklearn.model_selection import cross_val_score
scores = cross_val_score(model, X, y, cv=5)
print("Scores:", scores)
print("Mean:", scores.mean())
print("Standard deviation:", scores.std())

(reg, X, y, scoring="neg_mean_squared_error", cv=10)
のように適応度関数を指定することもできる。これはk個のモデルの比較に使うスコアである。

グリッドサーチ

from sklearn.model_selection import GridSearchCV
param_grid = { : [ ]}
model = 
grid_search = GridSearchCV(estimator=model, param_grid=param_grid, cv=5)
grid_search,fit(X, y)
best_model = grid_search.best_estimator_
best_model.predict(X)
print(grid_search.best_params_)
print(grid_search.best_score_)