1、智能數(shù)據(jù)的劃分和介紹數(shù)據(jù)集1. 數(shù)據(jù)集劃分機(jī)器學(xué)習(xí)般的數(shù)據(jù)集會(huì)劃分為兩個(gè)部分訓(xùn)練數(shù)據(jù)：于訓(xùn)練，構(gòu)建模型測試數(shù)據(jù)：在模型檢驗(yàn)時(shí)使，于評(píng)估模型是否有效2. sklearn數(shù)據(jù)集接介紹3. sklearn分類數(shù)據(jù)集4. sklearn回歸數(shù)據(jù)集數(shù)據(jù)集劃分APIsklearn.model_selection.tranid_text_split數(shù)據(jù)集API介紹skiearn.datasets加載獲取流數(shù)據(jù)集datasets.load_*()獲取規(guī)模數(shù)據(jù)集，數(shù)據(jù)包含在datasetsdatasets.fetch_*(data_home=None)獲取規(guī)模數(shù)據(jù)集，需要從絡(luò)下載，函數(shù)第個(gè)參數(shù)是data_hom

2、e，表數(shù)據(jù)集下載的錄默認(rèn)是/scikit_learn_data/獲取數(shù)據(jù)集返回的類型load和fetch返回的數(shù)據(jù)類型datasets.base.Bunch(字典格式)data:特征數(shù)據(jù)數(shù)組，是n_samples*n_features的維numpy.ndarray數(shù)組targets：標(biāo)簽數(shù)組，是n_samples的維numpy.adarray數(shù)組DESCR：數(shù)據(jù)描述feature_names：特征名，新聞數(shù)據(jù)，寫數(shù)字，回歸數(shù)據(jù)集沒有target_names:標(biāo)簽名獲取數(shù)據(jù)集案例load_datas()if _name_ =_main_:main()運(yùn)結(jié)果：數(shù)據(jù)集進(jìn)分割sklearn.model

3、_selection.train_test_split(*arrays,*options)x 數(shù)據(jù)集的特征值y 數(shù)據(jù)集的標(biāo)簽值text_size 測試集的，般為floatrandom_state 隨機(jī)數(shù)種，不同的種會(huì)造成不同的隨機(jī)采樣結(jié)果，相同的種采樣結(jié)果相同return 訓(xùn)練集特征值，測試集特征值，訓(xùn)練標(biāo)簽，測試標(biāo)簽（默認(rèn)隨機(jī)?。┐a：liload_iris()print(.data)def data_split():irload_iris()#訓(xùn)練集特征值：x_train 測試值：y_train 測試集：特征值：x_text 訓(xùn)練集y_trainx_train,x_text,y_train

4、,y_texttrain_test_split(ir.data,ir.target,test_size0.25)print(,x_train,y_train)def main():data_split()if _name_ =_main_:main()運(yùn)結(jié)果：于分類的數(shù)據(jù)集（新聞）sklearn.datasets.fetch_20newsgroups(data_home=None,subset=“train”)subset:train“或者text,“all可選，選擇要加載的數(shù)據(jù)集，訓(xùn)練集的”訓(xùn)練”,測試集的“測試”,兩者的“全部”datasets.clear_data_home(data_h

5、ome=None)清除錄下的數(shù)據(jù)代碼：liload_iris()print(.data)def data_split():irload_iris()#訓(xùn)練集特征值：x_train 測試值：y_train 測試集：特征值：x_text 訓(xùn)練集y_trainx_train,x_text,y_train,y_texttrain_test_split(ir.data,ir.target,test_size0.25)print(,x_train,y_train)def get_news():newsfetch_20newsgroups(subset=all)print(news.data)if _name_ =_main_:main()運(yùn)結(jié)果：回歸數(shù)據(jù)集代碼：liload_iris()def data_split():irload_iris()#訓(xùn)練集特征值：x_train 測試值：y_train 測試集：特征值：x_text 訓(xùn)練集y_trainx_train,x_text,y_train,y_texttrain_test_split(ir.data,ir.target,test_size0.25)print(,x_train,y_train)def get_news():newsfetch_20newsgroups(subset=