推薦 ：21 句話入門機(jī)器學(xué)習(xí)！

>>> import numpy as np>>> members = np.array([  ['男', '25', 185, 80, '程序員', 35, 200,  30],  ['女', '23', 170, 55, '公務(wù)員', 15,   0,  80],  ['男', '30', 180, 82, '律師',   60, 260, 300],  ['女', '27', 168, 52, '記者',   20, 180, 150]])

>>> security = np.float32((members[:,-1])) # 提取有價(jià)證券特征列數(shù)據(jù)>>> securityarray([ 30.,  80., 300., 150.], dtype=float32)>>> (security - security.mean())/security.std() # 減去均值再除以標(biāo)準(zhǔn)差array([-1.081241, -0.5897678, 1.5727142, 0.09829464], dtype=float32)

>>> security = np.float32((members[:,-1])) # 提取有價(jià)證券特征列數(shù)據(jù)>>> securityarray([ 30.,  80., 300., 150.], dtype=float32)>>> (security - security.min())/(security.max() - security.min()) # 減去最小值再除以極差array([0., 0.18518518, 1., 0.44444445], dtype=float32)

>>> from sklearn import preprocessing as pp>>> X = [  ['男', '程序員'],  ['女', '公務(wù)員'],  ['男', '律師', ],  ['女', '記者', ]]>>> ohe = pp.OneHotEncoder().fit(X)>>> ohe.transform(X).toarray()array([[0., 1., 0., 0., 1., 0.],       [1., 0., 1., 0., 0., 0.],       [0., 1., 0., 1., 0., 0.],       [1., 0., 0., 0., 0., 1.]])

datasets.load_boston([return_X_y]) ：加載波士頓房?jī)r(jià)數(shù)據(jù)集datasets.load_breast_cancer([return_X_y]) ：加載威斯康星州乳腺癌數(shù)據(jù)集datasets.load_diabetes([return_X_y]) ：加載糖尿病數(shù)據(jù)集datasets.load_digits([n_class, return_X_y]) ：加載數(shù)字?jǐn)?shù)據(jù)集datasets.load_iris([return_X_y]) ：加載鳶尾花數(shù)據(jù)集。datasets.load_linnerud([return_X_y]) ：加載體能訓(xùn)練數(shù)據(jù)集datasets.load_wine([return_X_y]) ：加載葡萄酒數(shù)據(jù)集datasets.fetch_20newsgroups([data_home, …]) ：加載新聞文本分類數(shù)據(jù)集datasets.fetch_20newsgroups_vectorized([…]) ：加載新聞文本向量化數(shù)據(jù)集datasets.fetch_california_housing([…]) ：加載加利福尼亞住房數(shù)據(jù)集datasets.fetch_covtype([data_home, …]) ：加載森林植被數(shù)據(jù)集datasets.fetch_kddcup99([subset, data_home, …]) ：加載網(wǎng)絡(luò)入侵檢測(cè)數(shù)據(jù)集datasets.fetch_lfw_pairs([subset, …]) ：加載人臉（成對(duì)）數(shù)據(jù)集datasets.fetch_lfw_people([data_home, …]) ：加載人臉（帶標(biāo)簽）數(shù)據(jù)集datasets.fetch_olivetti_faces([data_home, …]) ：加載 Olivetti 人臉數(shù)據(jù)集datasets.fetch_rcv1([data_home, subset, …])：加載路透社英文新聞文本分類數(shù)據(jù)集datasets.fetch_species_distributions([…]) ：加載物種分布數(shù)據(jù)集

>>> from sklearn.datasets import load_iris>>> X, y = load_iris(return_X_y=True)>>> X.shape # 數(shù)據(jù)集X有150個(gè)樣本，4個(gè)特征列(150, 4)>>> y.shape # 標(biāo)簽集y的每一個(gè)標(biāo)簽和數(shù)據(jù)集X的每一個(gè)樣本一一對(duì)應(yīng)(150,)>>> X[0], y[0](array([5.1, 3.5, 1.4, 0.2]), 0)

>>> iris = load_iris()>>> iris.target_names # 查看標(biāo)簽的名字array(['setosa', 'versicolor', 'virginica'], dtype='<U10')>>> X = iris.data>>> y = iris.target

>>> from sklearn.datasets import load_iris>>> from sklearn.model_selection import train_test_split as tsplit>>> X, y = load_iris(return_X_y=True)>>> X_train, X_test, y_train, y_test = tsplit(X, y, test_size=0.1)>>> X_train.shape, X_test.shape((135, 4), (15, 4))>>> y_train.shape, y_test.shape((135,), (15,))

>>> from sklearn.datasets import load_iris>>> from sklearn.model_selection import train_test_split as tsplit>>> from sklearn.neighbors import KNeighborsClassifier # 導(dǎo)入k-近鄰分類模型>>> X, y = load_iris(return_X_y=True) # 獲取鳶尾花數(shù)據(jù)集，返回樣本集和標(biāo)簽集>>> X_train, X_test, y_train, y_test = tsplit(X, y, test_size=0.1) # 拆分為訓(xùn)練集和測(cè)試集>>> m = KNeighborsClassifier(n_neighbors=10) # 模型實(shí)例化，n_neighbors參數(shù)指定k值，默認(rèn)k=5>>> m.fit(X_train, y_train) # 模型訓(xùn)練KNeighborsClassifier()>>> m.predict(X_test) # 對(duì)測(cè)試集分類array([2, 1, 2, 2, 1, 2, 1, 2, 2, 1, 0, 1, 0, 0, 2])>>> y_test # 這是實(shí)際的分類情況，上面的預(yù)測(cè)只錯(cuò)了一個(gè)array([2, 1, 2, 2, 2, 2, 1, 2, 2, 1, 0, 1, 0, 0, 2])>>> m.score(X_test, y_test) # 模型測(cè)試精度（介于0~1）0.9333333333333333

>>> from sklearn.datasets import load_boston>>> from sklearn.model_selection import train_test_split as tsplit>>> from sklearn.neighbors import KNeighborsRegressor>>> X, y = load_boston(return_X_y=True) # 加載波士頓房?jī)r(jià)數(shù)據(jù)集>>> X.shape, y.shape, y.dtype # 該數(shù)據(jù)集共有506個(gè)樣本，13個(gè)特征列，標(biāo)簽集為浮點(diǎn)型，適用于回歸模型((506, 13), (506,), dtype('float64'))>>> X_train, X_test, y_train, y_test = tsplit(X, y, test_size=0.01) # 拆分為訓(xùn)練集和測(cè)試集>>> m = KNeighborsRegressor(n_neighbors=10) # 模型實(shí)例化，n_neighbors參數(shù)指定k值，默認(rèn)k=5>>> m.fit(X_train, y_train) # 模型訓(xùn)練KNeighborsRegressor(n_neighbors=10)>>> m.predict(X_test) # 預(yù)測(cè)6個(gè)測(cè)試樣本的房?jī)r(jià)array([27.15, 31.97, 12.68, 28.52, 20.59, 21.47])>>> y_test # 這是測(cè)試樣本的實(shí)際價(jià)格，除了第2個(gè)（索引為1）樣本偏差較大，其他樣本偏差還算差強(qiáng)人意array([29.1, 50. , 12.7, 22.8, 20.4, 21.5])

>>> from sklearn import metrics>>> y_pred = m.predict(X_test)>>> metrics.mean_squared_error(y_test, y_pred) # 均方誤差60.27319999999995>>> metrics.median_absolute_error(y_test, y_pred) # 中位數(shù)絕對(duì)誤差1.0700000000000003>>> metrics.r2_score(y_test, y_pred) # 復(fù)相關(guān)系數(shù)0.5612816401629652

>>> from sklearn.datasets import load_boston>>> from sklearn.model_selection import train_test_split as tsplit>>> from sklearn.tree import DecisionTreeRegressor>>> X, y = load_boston(return_X_y=True) # 加載波士頓房?jī)r(jià)數(shù)據(jù)集>>> X_train, X_test, y_train, y_test = tsplit(X, y, test_size=0.01) # 拆分為訓(xùn)練集和測(cè)試集>>> m = DecisionTreeRegressor(max_depth=10) # 實(shí)例化模型，決策樹(shù)深度為10>>> m.fit(X, y) # 訓(xùn)練DecisionTreeRegressor(max_depth=10)>>> y_pred = m.predict(X_test) # 預(yù)測(cè)>>> y_test # 這是測(cè)試樣本的實(shí)際價(jià)格，除了第2個(gè)（索引為1）樣本偏差略大，其他樣本偏差較小array([20.4,  21.9,  13.8,  22.4, 13.1,  7. ])>>> y_pred # 這是6個(gè)測(cè)試樣本的預(yù)測(cè)房?jī)r(jià)，非常接近實(shí)際價(jià)格array([20.14, 22.33, 14.34, 22.4, 14.62, 7. ])>>> metrics.r2_score(y_test, y_pred) # 復(fù)相關(guān)系數(shù)0.9848774474870712>>> metrics.mean_squared_error(y_test, y_pred) # 均方誤差0.4744784865112032>>> metrics.median_absolute_error(y_test, y_pred) # 中位數(shù)絕對(duì)誤差0.3462962962962983

>>> from sklearn.datasets import load_diabetes>>> from sklearn.model_selection import train_test_split as tsplit>>> from sklearn.svm import SVR>>> from sklearn import metrics>>> X, y = load_diabetes(return_X_y=True)>>> X.shape, y.shape, y.dtype((442, 10), (442,), dtype('float64'))>>> X_train, X_test, y_train, y_test = tsplit(X, y, test_size=0.02)>>> svr_1 = SVR(kernel='rbf', C=0.1) # 實(shí)例化SVR模型，rbf核函數(shù)，C=0.1>>> svr_2 = SVR(kernel='rbf', C=100) # 實(shí)例化SVR模型，rbf核函數(shù)，C=100>>> svr_1.fit(X_train, y_train) # 模型訓(xùn)練SVR(C=0.1)>>> svr_2.fit(X_train, y_train) # 模型訓(xùn)練SVR(C=100)>>> z_1 = svr_1.predict(X_test) # 模型預(yù)測(cè)>>> z_2 = svr_2.predict(X_test) # 模型預(yù)測(cè)>>> y_test # 這是測(cè)試集的實(shí)際值array([ 49., 317.,  84., 181., 281., 198.,  84.,  52., 129.])>>> z_1 # 這是C=0.1的預(yù)測(cè)值，偏差很大array([138.10720127, 142.1545034 , 141.25165838, 142.28652449,       143.19648143, 143.24670732, 137.57932272, 140.51891989,       143.24486911])>>> z_2 # 這是C=100的預(yù)測(cè)值，偏差明顯變小array([ 54.38891948, 264.1433666 , 169.71195204, 177.28782561,       283.65199575, 196.53405477,  61.31486045, 199.30275061,       184.94923477])>>> metrics.mean_squared_error(y_test, z_1) # C=0.01的均方誤差8464.946517460194>>> metrics.mean_squared_error(y_test, z_2) # C=100的均方誤差3948.37754995066>>> metrics.r2_score(y_test, z_1) # C=0.01的復(fù)相關(guān)系數(shù)0.013199351909129464>>> metrics.r2_score(y_test, z_2) # C=100的復(fù)相關(guān)系數(shù)0.5397181166871942>>> metrics.median_absolute_error(y_test, z_1) # C=0.01的中位數(shù)絕對(duì)誤差57.25165837797314>>> metrics.median_absolute_error(y_test, z_2) # C=100的中位數(shù)絕對(duì)誤差22.68513954888364

>>> from sklearn.datasets import load_breast_cancer # 導(dǎo)入數(shù)據(jù)加載函數(shù)>>> from sklearn.tree import DecisionTreeClassifier # 導(dǎo)入隨機(jī)樹(shù)>>> from sklearn.ensemble import RandomForestClassifier # 導(dǎo)入隨機(jī)森林>>> from sklearn.model_selection import cross_val_score # 導(dǎo)入交叉驗(yàn)證>>> ds = load_breast_cancer() # 加載威斯康星州乳腺癌數(shù)據(jù)集>>> ds.data.shape # 569個(gè)乳腺癌樣本，每個(gè)樣本包含30個(gè)特征(569, 30)>>> dtc = DecisionTreeClassifier() # 實(shí)例化決策樹(shù)分類模型>>> rfc = RandomForestClassifier() # 實(shí)例化隨機(jī)森林分類模型>>> dtc_scroe = cross_val_score(dtc, ds.data, ds.target, cv=10) # 交叉驗(yàn)證>>> dtc_scroe # 決策樹(shù)分類模型交叉驗(yàn)證10次的結(jié)果array([0.92982456, 0.85964912, 0.92982456, 0.89473684, 0.92982456,       0.89473684, 0.87719298, 0.94736842, 0.92982456, 0.92857143])>>> dtc_scroe.mean() # 決策樹(shù)分類模型交叉驗(yàn)證10次的平均精度0.9121553884711779>>> rfc_scroe = cross_val_score(rfc, ds.data, ds.target, cv=10) # 交叉驗(yàn)證>>> rfc_scroe # 隨機(jī)森林分類模型交叉驗(yàn)證10次的結(jié)果array([0.98245614, 0.89473684, 0.94736842, 0.94736842, 0.98245614,       0.98245614, 0.94736842, 0.98245614, 0.94736842, 1.        ])>>> rfc_scroe.mean()# 隨機(jī)森林分類模型交叉驗(yàn)證10次的平均精度0.9614035087719298

    
    

     >>> from sklearn import datasets as dss # 導(dǎo)入樣本生成器
    
    
>>> from sklearn.cluster import KMeans # 從聚類子模塊導(dǎo)入聚類模型>>> import matplotlib.pyplot as plt>>> plt.rcParams['font.sans-serif'] = ['FangSong']>>> plt.rcParams['axes.unicode_minus'] = False>>> X_blob, y_blob = dss.make_blobs(n_samples=[300,400,300], n_features=2)>>> X_circle, y_circle = dss.make_circles(n_samples=1000, noise=0.05, factor=0.5)>>> X_moon, y_moon = dss.make_moons(n_samples=1000, noise=0.05)>>> y_blob_pred = KMeans(init='k-means++', n_clusters=3).fit_predict(X_blob)>>> y_circle_pred = KMeans(init='k-means++', n_clusters=2).fit_predict(X_circle)>>> y_moon_pred = KMeans(init='k-means++', n_clusters=2).fit_predict(X_moon)>>> plt.subplot(131)<matplotlib.axes._subplots.AxesSubplot object at 0x00000180AFDECB88>>>> plt.title('團(tuán)狀簇')Text(0.5, 1.0, '團(tuán)狀簇')>>> plt.scatter(X_blob[:,0], X_blob[:,1], c=y_blob_pred)<matplotlib.collections.PathCollection object at 0x00000180C495DF08>>>> plt.subplot(132)<matplotlib.axes._subplots.AxesSubplot object at 0x00000180C493FA08>>>> plt.title('環(huán)狀簇')Text(0.5, 1.0, '環(huán)狀簇')>>> plt.scatter(X_circle[:,0], X_circle[:,1], c=y_circle_pred)<matplotlib.collections.PathCollection object at 0x00000180C499B888>>>> plt.subplot(133)<matplotlib.axes._subplots.AxesSubplot object at 0x00000180C4981188>>>> plt.title('新月簇')Text(0.5, 1.0, '新月簇')>>> plt.scatter(X_moon[:,0], X_moon[:,1], c=y_moon_pred)<matplotlib.collections.PathCollection object at 0x00000180C49DD1C8>>>> plt.show()

>>> from sklearn import datasets as dss>>> from sklearn.cluster import DBSCAN>>> import matplotlib.pyplot as plt>>> plt.rcParams['font.sans-serif'] = ['FangSong']>>> plt.rcParams['axes.unicode_minus'] = False>>> X, y = dss.make_moons(n_samples=1000, noise=0.05)>>> dbs_1 = DBSCAN() # 默認(rèn)核心樣本半徑0.5，核心樣本鄰居5個(gè)>>> dbs_2 = DBSCAN(eps=0.2) # 核心樣本半徑0.2，核心樣本鄰居5個(gè)>>> dbs_3 = DBSCAN(eps=0.1) # 核心樣本半徑0.1，核心樣本鄰居5個(gè)>>> dbs_1.fit(X)DBSCAN(algorithm='auto', eps=0.5, leaf_size=30, metric='euclidean', metric_params=None, min_samples=5, n_jobs=None, p=None)>>> dbs_2.fit(X)DBSCAN(algorithm='auto', eps=0.2, leaf_size=30, metric='euclidean', metric_params=None, min_samples=5, n_jobs=None, p=None)>>> dbs_3.fit(X)DBSCAN(algorithm='auto', eps=0.1, leaf_size=30, metric='euclidean', metric_params=None, min_samples=5, n_jobs=None, p=None)>>> plt.subplot(131)<matplotlib.axes._subplots.AxesSubplot object at 0x00000180C4C5D708>>>> plt.title('eps=0.5')Text(0.5, 1.0, 'eps=0.5')>>> plt.scatter(X[:,0], X[:,1], c=dbs_1.labels_)<matplotlib.collections.PathCollection object at 0x00000180C4C46348>>>> plt.subplot(132)<matplotlib.axes._subplots.AxesSubplot object at 0x00000180C4C462C8>>>> plt.title('eps=0.2')Text(0.5, 1.0, 'eps=0.2')>>> plt.scatter(X[:,0], X[:,1], c=dbs_2.labels_)<matplotlib.collections.PathCollection object at 0x00000180C49FC8C8>>>> plt.subplot(133)<matplotlib.axes._subplots.AxesSubplot object at 0x00000180C49FCC08>>>> plt.title('eps=0.1')Text(0.5, 1.0, 'eps=0.1')>>> plt.scatter(X[:,0], X[:,1], c=dbs_3.labels_)<matplotlib.collections.PathCollection object at 0x00000180C49FC4C8>>>> plt.show()

>>> from sklearn import datasets as dss>>> from sklearn.decomposition import PCA>>> ds = dss.load_iris()>>> ds.data.shape # 150個(gè)樣本，4個(gè)特征維(150, 4)>>> m = PCA() # 使用默認(rèn)參數(shù)實(shí)例化PCA類，n_components=None>>> m.fit(ds.data)PCA(copy=True, iterated_power='auto', n_components=None, random_state=None, svd_solver='auto', tol=0.0, whiten=False)>>> m.explained_variance_ # 正交變換后各成分的方差值array([4.22824171, 0.24267075, 0.0782095 , 0.02383509])>>> m.explained_variance_ratio_ # 正交變換后各成分的方差值占總方差值的比例array([0.92461872, 0.05306648, 0.01710261, 0.00521218])

>>> m = PCA(n_components=0.97)>>> m.fit(ds.data)PCA(copy=True, iterated_power='auto', n_components=0.97, random_state=None, svd_solver='auto', tol=0.0, whiten=False)>>> m.explained_variance_array([4.22824171, 0.24267075])>>> m.explained_variance_ratio_array([0.92461872, 0.05306648])>>> d = m.transform(ds.data)>>> d.shape(150, 2)

>>> import matplotlib.pyplot as plt>>> plt.scatter(d[:,0], d[:,1], c=ds.target)<matplotlib.collections.PathCollection object at 0x0000016FBF243CC8>>>> plt.show()