机器学习之决策树在sklearn中的实现.docx

《机器学习之决策树在sklearn中的实现.docx》由会员分享，可在线阅读，更多相关《机器学习之决策树在sklearn中的实现.docx（8页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、I JU.HU:iiDpicrQQaiatriiiaiypRYtaiooroiaiioaiai9biRcipiJ， (；：!WWwWloaii 八、sr,g”， I =*B!9?iWR1小伙伴们大家好。（一一）7、,首先声明一下，我的开发环境是JupyteMab,所用的库和版本 大家参考：Python 3.7.1 （你的版本至少要3.4以上Scikit-learn 0.20.0 （你的版本至少要 0.20Graphviz 0.8.4 （没有画不出决策树哦，安装代码 conda install python-graphviz用SKIearn建立一棵决策树这里采纳的数据集是SKIearn中的红酒数

2、据集。1导入需要的算法库和模块from skleam import tree #导入 tree模块from sklearn.datasets import load_wine #导入红酒数据集from sklearn.model_selection import train_test_split #导入训I 练集和测试集切分包v/pre2探究数据wine = load_wine()wine.datawine.targetwine.target.shape运行的结果是这样子的:In 2:wi ne = load_wine()wi ne . dataarray(1.423e+01, 1.710e+

3、00,l.O65e+03,1.320e+01, 1.780e+00,1.050e+03,1.316e+01, 2.36Oe+0O, 1.185e+03, ,1.327e+01, 4.280e+O0, 8.350e+02,1.317e+01, 2.590e+00, 8.400e+O2,1.413e+01, 4.10Oe+OO, 5.60Oe+O2)array(1.423e+01, 1.710e+00,l.O65e+03,1.320e+01, 1.780e+00,1.050e+03,1.316e+01, 2.36Oe+0O, 1.185e+03, ,1.327e+01, 4.280e+O0, 8.

4、350e+02,1.317e+01, 2.590e+00, 8.400e+O2,1.413e+01, 4.10Oe+OO, 5.60Oe+O2)2.430e+00, 1.040e+00, 3.920e+00,2.140e+00,1.050e+00, 3.400e+00,2.670e+00,1.030e+0O, 3.170e+00,2.260e+00,5.900e-01, 1.560e+00,2.370e+00, 6000e-01, 1.620e+00,2.740e+00,6.100e-01, 1.600e+00,wine.data.shape (178, 13)data就是该数据集的特征矩阵，

5、从运行结果可以看出，该红酒数据集一共有178条纪录，13 个特征。wine.data.shape (178, 13)wi ne , targetOut 5:array(0,3330,30,0,0,30,30,30,30,*33330,0,333333333333330,0,0,0,0,31,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,

6、2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 2)wi ne , target.shapeOut6: (178,)特征矩阵中有178条纪录，相对应的标签Y就有178个数据。假如wine是一张表，应当长这样:import pandas as pd pd.concat(pd.Datafrxxxxame(wine.data)pd.Datafrxxxxame(wine.target)axis=l)In 11:In 12：Out12：import pandas as pdpd.concat(p

7、d , DataFrame(wine.data),pd.DataFrame(wine.target),axis=l)01234567891011120014.231.712.4315.6127.02.803.060.282.295.6400001.043.921065.00113.201.782.1411.2100.02.652.760.261.2843800001.053.401050.00213.162.362.6718.6101.02.803.240.302.815.6800001.033.171185.00314371.952.5016.8113.03.853.490.242.187.

8、8000000.863.451480.00413.242.592.8721.0118.02.802.690.391.8243200001.042.93735.00514.201.762.4S15.2112.03.273.390341.976.7500001.052.851450.00614.391.872.4514.696.02.502.520301.985.2500001.023.581290.00714.062.152.6117.6121.02.602.510.311.255.O5OOOO1.063.581295.00814.831.642.1714.097.02.802.980291.9

9、85.2000001.082.851045.00913.861.352.2716.098.02.983.150.221.857.2200001.013.551045.001014.102.162.3018.0105.02.953.320.222.385.7500001.253.171510.001114.121.4823216.895.02.202.430.261.575.0000001.172.821280.001213.751.732.4116.089.02.602.760.291.815.6000001.152.901320.001314.751.732.3911.491.03.103.

10、690.432.815.4000001.252.731150.001414.381.872.3812.0102.03.303.640.292.967.5000001.203.001547.001513.631.812.7017.2112.02.852.910301.4673000001.282.881310.001614.301.922.7220.0120.02.803.140.331.976.2000001.072.651280.001713.831.572.6220.0115.02.953.400.401.726.6000001.132.571130.001814.191.592.4816

11、.5108.03.303.930.321.868.7000G01.232.821680.001913.643.102.5615.2116.02.703.030.171.665.1000000.963.36845.002014.061.632.2816.0126.03.003.170.242.105.6500001.093.71780.002112.933.802.6518.6102.02.412.410.251.984.5000001.033.52770.002213.711.862.3616.6101.02.612.880.271.693.8000001.114.001035.002312.

12、851.602.5217.895.02.482.370.261.463.9300001.093.631015.002413.501.812.6120.096.02.532.610.281.663.52OOGO1.123.82845.002513.052.053.2225.0124.02.632.680.471.923.5800001.133.20830.002613.391.772.6216.193.02.852.940341.454.8000000.923.221195.002713.301.722.1417.094.02.402.190271.353.9500001.022.771285.

13、002813.871.902.8019.4107.02.952.970371764.5000G01.253.40915.002914.021.682.2116.096.02.652.330.261.984.7000001.043.591035.00 14813323.2423821.592.01.930.760.451.258.4200000.551.62650.0214913.083.902.3621.5113.01.411.390341.149.4000000.571.33550.0215013.503.122.6224.0123.01.401.570.221.258.6000000.59

14、1.30500.0215112.792.672.4822.0112.01.481.360.241.2610.8000000.481.47480.0215213.111.902.7525.5116.02.201.280.261.567.1000000.611.33425.0215313.233.302.2818.598.01.800.830.611.8710.5200000.561.51675.0215412.581.292.1020.0103.01.480.580.531.407.6000000.581.55640.0215513.175.192.3222.093.01.740.630.611

15、.557.9000000.601.48725.0215613.844.1223819.589.01.800.830.481.569.0100000.571.64480.0215712.453.032.6427.097.01.900.580.631.147.5000000.671.73880.02158 1434 1.68 2.70 25.098.0 2.80 1.31 0.53 2.70 13.000000 0.57 1.96660.0这是数据集特征列名和标签分类wine.feature_names wine.target_nameswi ne , feature_namesOut13: *a

16、lcohol1,malic acid 1. 1 ash *,1alcali ni ty_of_ash,1magnesi um1,total_phenols1,flavanoids,nonflavano-id_phenols , ,1proanthocyani ns 1,color_-intensnty1 ,hue * ,od280/od315_of_diluted_wines,1proli neIn 14:wi ne target_namesOut14:array(1class_0, class_l1, * class_21, dtype=1U7 1)3分训练集和测试集这里选取30%作为测试集

17、。切分好之后，训练集有124条数据，测试集有54条数据。Xtrain Xtest Ytrain Ytest =train_test_split(wine.datawine.targettest_size=0.3)Xtrain.shapeXtest. shape15 : Xtrain, Xtest, Ytrain, Ytest = trann_test_split(wine.data,wine.target,test_size=0.3)In 16_: Xtrain.shapeOut16: (124, 13)In 171: Xtest. shapeOut17 : (54, 13)4建立模型clf

18、=tree.DecisionTreeClassifier(criterion=,entropyH) # 初始化树模型clf=clf.fit(Xtrain Ytrain) #实例化训练集score = clf.score(Xtest Ytest) #返回猜测的精确 度scoreelf = tree.DecisionTreeClassifier(criterion=,entropyn) #初始化树模型elf = elf. fitCXtrain, Ytrain)#实例化训练集score = elf.score(Xtest, Ytest)#返回预测的准确度score0.8333333333333334

19、5画出一棵树吧feature_name =酒精苹果酸 灰恢的碱性镁总盼类黄酮啡黄烷类酚类花青素颜色强度色调od280/od315稀释葡萄酒” 脯氨酸import graphvizdot_data = tree.export_graphviz(clfout_file=Nonefeature_names= feature_nameclass_names=琴酒”雪莉,贝尔摩德Ifilled=Truerounded=True)graph = graphviz.Source(dot_data)graphimport graphviz feature_name =泗，卜，苹累酸灰J灰的诚件，铁，总后I类黄

20、酮，生黄烷类的类J花吉:顶色强度;色调，od28od315稀林葡荀酒二锦氨像 dot_data = tree.export graphviz(clf,out_file=None,feature_names= feature_name,class_nlmes=琴泗，贝尔佗”,filled=True,rounded=True)graph = graphviz.Source(dot_data)graph6探究决策树#特征重要性 clf.feature_importances_*zip(feature_nameclf.feature_importances_)19: #特征重要性elf.feature

21、_importances_array（0., 0,0,0,0,0., 0.14641189, 0, 0.，0.29304668,0.11825045, 0, 0.44229098）*zip（feature_nameclf.feature_importances_）20:（酒精L 0.0）,（苹果酸1 0.0）,（灰0.0）,（灰的碱性0.0）,（镁0.0）,（总酚 1 0.0）,（类黄酮非黄烷类酚类10.0）,（花青素1 0.0）,（颜色强度0.29304667859613337）,（色调（，od280/od315稀释葡萄酒0.0）, （脯氨酸 1 0.44229098082354273）到现在为止，我们已经学会建立一棵完整的决策树了。有爱好的话，动手建立一棵属于自己的决策树吧