我希望有一个流程,该流程可以为我提供一系列机器学习模型及其准确性得分,但仅针对能够提供该类型模型最佳结果的参数集。
例如,这里只是XGBoost的简历:
数据集:
import numpy as np
import pandas as pd
from sklearn.datasets import load_iris
iris = load_iris()
data = pd.DataFrame(data= np.c_[iris['data'], iris['target']],
columns= iris['feature_names'] + ['target'])
from sklearn.model_selection import train_test_split
X = data.drop(['target'], axis=1)
y = data['target']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=42)
查找最佳参数的功能:
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import accuracy_score, make_scorer
accu = make_scorer(accuracy_score) # I will be using f1 in future
def predict_for_best_params(alg, X_train, y_train, X_test):
params = {'n_estimators': [200, 300, 500]}
clf = GridSearchCV(alg, params, scoring = accu, cv=2)
clf.fit(X_train, y_train)
print(clf.best_estimator_)
y_pred = clf.predict(X_test)
return y_pred
在一种模型上使用它:
from xgboost import XGBClassifier
alg = [XGBClassifier()]
y_pred = predict_for_best_params(alg[0], X_train, y_train, X_test)
from sklearn.metrics import accuracy_score
print(accuracy_score(y_test, y_pred))
我想要实现的目标是:
from xgboost import XGBClassifier
from sklearn.ensemble import RandomForestClassifier
alg = [XGBClassifier(), RandomForrest()] # list of many of them
alg_params = {'XGBClassifier': [{'n_estimators': [200, 300, 500]}],
'RandomForrest': [{'max_depth ': [1, 2, 3, 4]}]}
def predict_for_best_params(alg, X_train, y_train, X_test, params):
clf = GridSearchCV(alg, params, scoring = accu, cv=2)
clf.fit(X_train, y_train)
print(clf.best_estimator_)
y_pred = clf.predict(X_test)
return y_pred
for algo in alg:
params = alg_params[str(algo)][0] #this won't work because str(algo) <> e.g. XGBClassifier() but XGBClassier(all default params)
y_pred = predict_for_best_params(algo, X_train, y_train, X_test, params)
print('{} accuracy is: {}'.format(algo, accuracy_score(y_test, y_pred)))
这是实现它的好方法吗?
答案 0 :(得分:2)
如果您只担心如何放置钥匙,则可以使用
params = alg_params[alg.__class__.__name__][0]
这应该仅返回alg
对象的类名
对于另一种方法,您可以看看我的其他答案:
该答案利用了GridSearchCV可以采用参数组合字典列表的事实,其中每个列表将单独扩展。但请注意以下几点:
for-loop
(使用多重处理),这可能比当前的n_jobs > 1
快。cv_results_
的{{1}}属性来分析分数。GridSearchCV
,您可以过滤y_pred
(也许通过将其导入cv_results_
),然后再次使用最合适的参数对估算器进行拟合,然后然后计算y_pred。但是应该很容易。