網(wǎng)格搜索 隨機(jī)搜索

• 網(wǎng)格搜索
• 隨機(jī)搜索

網(wǎng)格搜索

class sklearn.model_selection.GridSearchCV(

estimator :考慮優(yōu)化的估計(jì)器

param_grid : dict or list of dictionaries，希望進(jìn)行搜索的參數(shù)陣

scoring = None : string/callable/list/tuple/dict/None，模型評分方法

fit_params = None, n_jobs = 1, cv = None

iid = True :數(shù)據(jù)是否在各fold間均勻分布，此時(shí)將直接最小化總樣本的損失函數(shù)refit = True :是否使用發(fā)現(xiàn)的最佳參數(shù)重新擬合估計(jì)器

verbose = 0, pre_dispatch = '2*n_jobs ', error_score = 'raise’return_train_score = True :是否返回訓(xùn)練集的評分

)

GridSearchCV類的屬性:

cv_results_ :字典格式的參數(shù)列表，可被直接轉(zhuǎn)換為pandas數(shù)據(jù)框best_estimator_ :網(wǎng)格搜索得出的最佳模型best_score_ :最佳模型的平均交互驗(yàn)證得分best_params_ : dict，最佳模型的參數(shù)設(shè)定best_index_ : int，最佳模型對應(yīng)的索引值scorer_ : function or a dict，用于選擇最佳模型的評分函數(shù)n_splits_ : int，交叉驗(yàn)證的拆分?jǐn)?shù)

GridSearchCV類的方法:

decision_function ( *args，**kwargs) :調(diào)用篩選出的最佳模型并返回預(yù)測結(jié)果其余標(biāo)準(zhǔn)API接函數(shù)

decision_function返回?cái)?shù)據(jù)點(diǎn)屬于每個(gè)類別的判定系數(shù)，若為正數(shù)，則代表該點(diǎn)屬于這一類，負(fù)數(shù)則表示該點(diǎn)不屬于這一類。判定系數(shù)的絕對值越大，判斷的可信度越高。

from sklearn import svm from sklearn import datasets from sklearn.model_selection import GridSearchCV import pandas as pd iris = datasets.load_iris() parameters = {'kernel':('linear','rbf'),'C':[1.10]} #實(shí)例化 svc = svm.SVC(probability = True) clf = GridSearchCV(svc,parameters) clf.fit(iris.data,iris.target) GridSearchCV(cv='warn', error_score='raise-deprecating',estimator=SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,decision_function_shape='ovr', degree=3,gamma='auto_deprecated', kernel='rbf', max_iter=-1,probability=True, random_state=None, shrinking=True,tol=0.001, verbose=False),iid='warn', n_jobs=None,param_grid={'C': [1.1], 'kernel': ('linear', 'rbf')},pre_dispatch='2*n_jobs', refit=True, return_train_score=False,scoring=None, verbose=0) from sklearn.svm import SVC

#顯示所有擬合模型的參數(shù)設(shè)定 pd.DataFrame(clf.cv_results_) mean_fit_timestd_fit_timemean_score_timestd_score_timeparam_Cparam_kernelparamssplit0_test_scoresplit1_test_scoresplit2_test_scoremean_test_scorestd_test_scorerank_test_score01

0.001341	0.000485	0.000000	0.00000	1.1	linear	{'C': 1.1, 'kernel': 'linear'}	1.000000	0.960784	1.000000	0.986667	0.018577	1
0.005318	0.005421	0.000333	0.00047	1.1	rbf	{'C': 1.1, 'kernel': 'rbf'}	0.980392	0.960784	0.979167	0.973333	0.009021	2

clf.best_params_ {'C': 1.1, 'kernel': 'linear'} print(clf.decision_function(iris.data))#網(wǎng)格搜索 [[ 2.24627744 1.29829892 -0.30632837][ 2.23781119 1.29693571 -0.30471538][ 2.24548583 1.29718879 -0.30559412][ 2.23591041 1.29589056 -0.30389728][ 2.24795778 1.29822418 -0.3064484 ][ 2.23752685 1.29735806 -0.30495212][ 2.2434869 1.29624146 -0.30483534][ 2.24100113 1.29748846 -0.30534443][ 2.23661182 1.2953947 -0.30365738][ 2.2375786 1.2973143 -0.30492944][ 2.24558626 1.29895555 -0.30665596][ 2.23718064 1.29655761 -0.3044249 ][ 2.23997815 1.29729146 -0.30513039][ 2.2512266 1.29724082 -0.30622266][ 2.25842469 1.3013034 -0.30927775][ 2.25245732 1.29965175 -0.30772804][ 2.25194914 1.29891281 -0.3072508 ][ 2.24452273 1.29771579 -0.30581239][ 2.23723948 1.29851098 -0.30565066][ 2.24582594 1.29783236 -0.30600783][ 2.23172213 1.29768867 -0.30467686][ 2.24234659 1.29705619 -0.30520883][ 2.25994866 1.29879541 -0.30806837][ 2.22220568 1.29476653 -0.30201257][ 2.22238071 1.29523835 -0.30233811][ 2.22837826 1.29635472 -0.30354531][ 2.23246711 1.29580915 -0.30353915][ 2.24249366 1.29815377 -0.30588644][ 2.24452963 1.29837334 -0.30620727][ 2.23329426 1.2959152 -0.30367832][ 2.23097932 1.29600008 -0.30353221][ 2.23636238 1.29729451 -0.30481098][ 2.25330507 1.29966903 -0.30782355][ 2.2557716 1.30030318 -0.30843809][ 2.23545141 1.29669764 -0.30435806][ 2.2485707 1.29833029 -0.30657146][ 2.2491945 1.29959712 -0.30737741][ 2.24969301 1.29855531 -0.30681499][ 2.24251854 1.29603049 -0.30461341][ 2.24086415 1.29773868 -0.30548479][ 2.24814273 1.29786457 -0.3062567 ][ 2.22578517 1.29425389 -0.30197265][ 2.24585572 1.29640343 -0.30517211][ 2.22992334 1.29464894 -0.30258346][ 2.22639242 1.29551952 -0.3028421 ][ 2.23584715 1.29603506 -0.30398092][ 2.24398491 1.29802466 -0.30594642][ 2.24190045 1.29650998 -0.30483816][ 2.24570962 1.29872219 -0.30652718][ 2.24310378 1.29771639 -0.30567798][-0.25888157 2.27043006 0.84961887][-0.25420288 2.26613432 0.85736765][-0.26566202 2.26422889 1.03089525][-0.24725457 2.26389816 0.83953777][-0.26167944 2.26301954 0.97285293][-0.2558817 2.26139866 0.91727953][-0.25960377 2.26047526 0.98289875][-0.19807268 2.27437013 0.7462427 ][-0.25819096 2.26876747 0.85971716][-0.23803641 2.26291864 0.81574102][-0.22433167 2.27078106 0.76857795][-0.24705163 2.26470632 0.8338466 ][-0.24425962 2.27327002 0.78539706][-0.26147994 2.26048024 1.01994941][-0.21600922 2.27333011 0.75641414][-0.25140742 2.2712808 0.81140066][-0.2559883 2.25665492 0.98796188][-0.23905423 2.273164 0.77761803][-0.26533223 2.25720279 1.11452981][-0.23621673 2.27128811 0.78004053][-0.26495855 2.24680537 1.16854301][-0.23955415 2.2718155 0.78305262][-0.27060306 2.25321773 1.1785179 ][-0.26019837 2.26448894 0.92626119][-0.24921993 2.27097946 0.8065506 ][-0.25268911 2.26997762 0.82331433][-0.26531537 2.26507325 1.00569519][-0.27032769 2.25456052 1.17143878][-0.25759465 2.26029055 0.95342037][-0.20874231 2.27830251 0.74273904][-0.23361723 2.27122032 0.77690547][-0.22625653 2.27409475 0.76152715][-0.23467384 2.27203226 0.77563407][-0.27318245 1.24299275 2.21645134][-0.25579586 2.25427719 1.025428 ][-0.25255831 2.26139138 0.88790408][-0.2609764 2.26533568 0.92399971][-0.25977858 2.26537697 0.90938447][-0.23921988 2.26768065 0.79813843][-0.24402523 2.26517928 0.82109608][-0.2540305 2.26200534 0.89343164][-0.25783181 2.26278911 0.92107182][-0.24081212 2.27029624 0.79080786][-0.20208621 2.27458063 0.74730352][-0.2479946 2.26442341 0.83926823][-0.24139608 2.26915346 0.7964656 ][-0.24494709 2.26657115 0.81671407][-0.24899229 2.26941618 0.81397179][-0.16817711 2.27765932 0.73375002][-0.24293044 2.26735753 0.80760147][-0.28707384 1.14947354 2.28119096][-0.27544166 1.22551722 2.24241259][-0.28572654 1.21950229 2.26763227][-0.28075419 1.22722381 2.253935 ][-0.28496263 1.19919165 2.27167675][-0.29221928 1.20120343 2.28271138][-0.26350515 1.23390222 2.19566155][-0.28872304 1.22449799 2.27171205][-0.28536738 1.22270759 2.26580515][-0.28702396 1.1962758 2.27552215][-0.27295679 1.24123017 2.21846124][-0.27867286 1.23154175 2.24615515][-0.28099808 1.22753736 2.25431583][-0.27600114 1.21779044 2.24849707][-0.27838341 1.19196788 2.26253449][-0.27824419 1.21747936 2.25346649][-0.27892215 1.23592175 2.24334747][-0.29050259 1.20935777 2.27863355][-0.29632366 1.14630194 2.29280364][-0.27446017 1.24383688 2.21985224][-0.28353826 1.21366733 2.26542492][-0.27210994 1.22438478 2.23485951][-0.29332401 1.2025653 2.28418614][-0.2716128 1.24523345 2.20596604][-0.28175043 1.22264441 2.25843786][-0.28427267 1.23661991 2.25666445][-0.26887359 1.24715529 2.1886087 ][-0.2689827 1.24534071 2.19396581][-0.28314739 1.20961778 2.26596079][-0.28185127 1.24735998 2.24067838][-0.28776392 1.22705696 2.26896304][-0.2869051 1.23657142 2.26293051][-0.28371257 1.20202648 2.26897129][-0.27174181 1.25176897 2.18997946][-0.27988353 1.23805115 2.2441591 ][-0.2888502 1.21427543 2.27478929][-0.28160162 1.19629383 2.2668145 ][-0.27824449 1.23534175 2.24198858][-0.26693842 1.24624323 2.18133534][-0.27913621 1.23429921 2.24525582][-0.28339523 1.2030499 2.26818344][-0.27624857 1.23436839 2.23730158][-0.27544166 1.22551722 2.24241259][-0.28569412 1.20057856 2.27257418][-0.28405792 1.19376775 2.27133323][-0.27826174 1.22537803 2.24926502][-0.27556832 1.23623495 2.23346001][-0.27601105 1.23598869 2.23502853][-0.2782753 1.21132467 2.25621161][-0.27247876 1.23666418 2.22329881]]

隨機(jī)搜索

在不明確可能的參數(shù)候選值時(shí)，可以在指定的參數(shù)值分布中進(jìn)行取樣，實(shí)現(xiàn)對參數(shù)的隨機(jī)搜索。class sklearn.model_selection.RandomizedSearchCV(

estimator :param_distributions : dict，希望進(jìn)行搜索的參數(shù)字典n_iter = 10 : int，考慮抽取出的參數(shù)組合數(shù)，該參數(shù)用于控制計(jì)算總量scoring = None， fit _params = None, n_jobs = 1, iid = Truerefit = True :使用整個(gè)數(shù)據(jù)集重新fit搜索到的最佳參數(shù)組合模型cv = None, verbose =0 pre_dispatch = '2*n_jobs ', random_state = Noneerror_score = 'raise',return_train_score = True）

RandomizedSearchCV類的屬性（和GridSearchCV類相同):

cv_results_ :字典格式的參數(shù)列表，可被直接轉(zhuǎn)換為pandas數(shù)據(jù)框best_estimator_ : 網(wǎng)格搜索得出的最佳模型best_score_ :最佳模型的平均交互驗(yàn)證得分best_params_ : dict，最佳模型的參數(shù)設(shè)定best_index_ : int，最佳模型對應(yīng)的索引值scorer_ : function or a dict，用于選擇最佳模型的評分函數(shù)n_splits_ : int，交叉驗(yàn)證的拆分?jǐn)?shù)

RandomizedSearchCV類的方法（和GridSearchCV類相同)︰

decision_function (*args,**kwargs) :調(diào)用最佳模型，并返回預(yù)測結(jié)果其余標(biāo)準(zhǔn)API接函數(shù) import scipy.stats as stats from sklearn import datasets from sklearn.model_selection import RandomizedSearchCV import pandas as pd from sklearn.svm import SVC #導(dǎo)入數(shù)據(jù)集并設(shè)定參數(shù) iris = datasets.load_iris() parameters2 = {"kernel":('linear','rbf'),'C':stats.expon(scale=100),'gamma':stats.expon(scale=1),'class_weight':('balanced',None)} svc = SVC() clf = RandomizedSearchCV(svc,parameters2) clf.fit(iris.data,iris.target) pd.DataFrame(clf.cv_results_) mean_fit_timestd_fit_timemean_score_timestd_score_timeparam_Cparam_class_weightparam_gammaparam_kernelparamssplit0_test_scoresplit1_test_scoresplit2_test_scoremean_test_scorestd_test_scorerank_test_score0123456789

0.000998	8.142963e-04	0.000344	0.000486	133.412	balanced	1.61514	rbf	{'C': 133.41183918245417, 'class_weight': 'bal...	1.000000	0.941176	0.958333	0.966667	0.024918	6
0.000000	0.000000e+00	0.000653	0.000462	15.8825	None	1.69397	linear	{'C': 15.88249498314624, 'class_weight': None,...	1.000000	0.921569	1.000000	0.973333	0.037154	1
0.000665	4.710309e-04	0.000333	0.000470	37.7037	balanced	0.079273	linear	{'C': 37.703742462767295, 'class_weight': 'bal...	1.000000	0.921569	1.000000	0.973333	0.037154	1
0.000997	4.052337e-07	0.000333	0.000471	25.3494	balanced	0.951097	rbf	{'C': 25.34942982674127, 'class_weight': 'bala...	0.980392	0.941176	1.000000	0.973333	0.024415	1
0.000676	4.779710e-04	0.000343	0.000485	245.106	balanced	0.663588	linear	{'C': 245.10614694204088, 'class_weight': 'bal...	1.000000	0.921569	1.000000	0.973333	0.037154	1
0.000661	4.685966e-04	0.000337	0.000476	22.3682	None	0.0174971	linear	{'C': 22.368243827973412, 'class_weight': None...	1.000000	0.921569	1.000000	0.973333	0.037154	1
0.000333	4.703588e-04	0.000676	0.000478	227.165	None	0.318804	rbf	{'C': 227.16515187404892, 'class_weight': None...	0.980392	0.901961	1.000000	0.960000	0.042411	10
0.000332	4.691225e-04	0.000000	0.000000	125.798	None	0.11144	rbf	{'C': 125.79773872651727, 'class_weight': None...	1.000000	0.901961	1.000000	0.966667	0.046442	6
0.000000	0.000000e+00	0.000997	0.000027	3.60636	None	2.07116	linear	{'C': 3.6063649298660225, 'class_weight': None...	0.980392	0.921569	1.000000	0.966667	0.033333	6
0.000665	4.705275e-04	0.000322	0.000455	61.6049	None	0.429071	rbf	{'C': 61.60492358027872, 'class_weight': None,...	1.000000	0.901961	1.000000	0.966667	0.046442	6

clf.best_params_ {'C': 15.88249498314624,'class_weight': None,'gamma': 1.693966803275749,'kernel': 'linear'}

總結(jié)

以上是生活随笔為你收集整理的网格搜索、随机搜索机器学习实战的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。