Decission Tree<\/p>\n\n\n\n
\u521d\u59cb\u5316model<\/strong> \u8b93model\u5b78\u7fd2<\/strong> \u770bmodel\u5b78\u7fd2\u72c0\u6cc1<\/strong> \u6839\u64dainput\u9810\u6e2coutput<\/strong> refer ………………………………………………………………..<\/p>\n\n\n\n Install<\/strong> export_graphviz<\/strong> refer pydot<\/strong> …………………………………………………………………………..<\/p>\n\n\n\n #vi dataset<\/strong> #vi train.py<\/strong><\/p>\n\n\n\n <\/p>\n","protected":false},"excerpt":{"rendered":" Decission Tree \u8f09\u5165\u6a21\u7d44from sklear …<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"","fifu_image_alt":"","_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":false,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2}},"categories":[13],"tags":[],"class_list":["post-469","post","type-post","status-publish","format-standard","hentry","category-dataanalysis"],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"jetpack-related-posts":[{"id":465,"url":"https:\/\/systw.net\/note\/archives\/465","url_meta":{"origin":469,"position":0},"title":"SKLearn Logistic Regression","author":"raymond","date":"2019 \u5e74 9 \u6708 17 \u65e5","format":false,"excerpt":"Logistic Regression\u5c6c\u65bc\u5206\u985e\u6f14\u7b97\u6cd5\uff0c\u900f\u904eregression\u65b9\u5f0f\u9054\u5230\u5224\u65b7\u985e\u5225\u7684\u6548\u679c\u2026","rel":"","context":"\u5728\u300c\u6578\u64da\u5206\u6790\u300d\u4e2d","block_context":{"text":"\u6578\u64da\u5206\u6790","link":"https:\/\/systw.net\/note\/archives\/category\/dataanalysis"},"img":{"alt_text":"","src":"","width":0,"height":0},"classes":[]},{"id":467,"url":"https:\/\/systw.net\/note\/archives\/467","url_meta":{"origin":469,"position":1},"title":"SKLearn Linear Regression","author":"raymond","date":"2019 \u5e74 4 \u6708 27 \u65e5","format":false,"excerpt":"\u4f7f\u7528OLS( Ordinary least squares ) \u7684Linear Regression\u2026","rel":"","context":"\u5728\u300c\u6578\u64da\u5206\u6790\u300d\u4e2d","block_context":{"text":"\u6578\u64da\u5206\u6790","link":"https:\/\/systw.net\/note\/archives\/category\/dataanalysis"},"img":{"alt_text":"","src":"","width":0,"height":0},"classes":[]},{"id":492,"url":"https:\/\/systw.net\/note\/archives\/492","url_meta":{"origin":469,"position":2},"title":"Decision Tree","author":"raymond","date":"2012 \u5e74 11 \u6708 29 \u65e5","format":false,"excerpt":"\u6c7a\u7b56\u6a39\u61c9\u7528:\u3000\u5e38\u7528\u5728\u5206\u985e,\u4e5f\u5c31\u662f\u5148\u5f9e\u8cc7\u6599\u4e2d\u627e\u51fa\u898f\u5247,\u5728\u5229\u7528\u8a72\u898f\u5247\u5f9e\u5176\u4ed6\u8cc7\u6599\u4e2d\u627e\u51fa\u7d50\u679c\u00a0\u5e38\u898b\u2026","rel":"","context":"\u5728\u300c\u6578\u64da\u5206\u6790\u300d\u4e2d","block_context":{"text":"\u6578\u64da\u5206\u6790","link":"https:\/\/systw.net\/note\/archives\/category\/dataanalysis"},"img":{"alt_text":"","src":"","width":0,"height":0},"classes":[]},{"id":473,"url":"https:\/\/systw.net\/note\/archives\/473","url_meta":{"origin":469,"position":3},"title":"SKLearn Dataset","author":"raymond","date":"2019 \u5e74 7 \u6708 12 \u65e5","format":false,"excerpt":"\u8cc7\u6599\u96c6\u7684\u4f86\u6e90\u4e3b\u8981\u6709\u5169\u7a2e\uff0c\u5167\u5efa\u8cc7\u6599\u96c6\uff0c\u6216\u662f\u7531\u5916\u90e8\u8f09\u5165\u7684\u8cc7\u6599\u96c6 ....................\u2026","rel":"","context":"\u5728\u300c\u6578\u64da\u5206\u6790\u300d\u4e2d","block_context":{"text":"\u6578\u64da\u5206\u6790","link":"https:\/\/systw.net\/note\/archives\/category\/dataanalysis"},"img":{"alt_text":"","src":"","width":0,"height":0},"classes":[]},{"id":486,"url":"https:\/\/systw.net\/note\/archives\/486","url_meta":{"origin":469,"position":4},"title":"Mahout Logistic","author":"raymond","date":"2014 \u5e74 9 \u6708 11 \u65e5","format":false,"excerpt":"mahout logistic regressionmahout\u7684logistic regressi\u2026","rel":"","context":"\u5728\u300c\u6578\u64da\u5206\u6790\u300d\u4e2d","block_context":{"text":"\u6578\u64da\u5206\u6790","link":"https:\/\/systw.net\/note\/archives\/category\/dataanalysis"},"img":{"alt_text":"","src":"","width":0,"height":0},"classes":[]},{"id":490,"url":"https:\/\/systw.net\/note\/archives\/490","url_meta":{"origin":469,"position":5},"title":"Neural Network","author":"raymond","date":"2012 \u5e74 12 \u6708 30 \u65e5","format":false,"excerpt":"Artificial Neural Network(\u985e\u795e\u7d93\u7db2\u8def)\u65bc1987\u5e74\u5f8c\u958b\u59cb\u666e\u53ca\u53ef\u5206\u70baSupe\u2026","rel":"","context":"\u5728\u300c\u6578\u64da\u5206\u6790\u300d\u4e2d","block_context":{"text":"\u6578\u64da\u5206\u6790","link":"https:\/\/systw.net\/note\/archives\/category\/dataanalysis"},"img":{"alt_text":"","src":"","width":0,"height":0},"classes":[]}],"_links":{"self":[{"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/posts\/469","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/comments?post=469"}],"version-history":[{"count":0,"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/posts\/469\/revisions"}],"wp:attachment":[{"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/media?parent=469"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/categories?post=469"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/systw.net\/note\/wp-json\/wp\/v2\/tags?post=469"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\u8f09\u5165\u6a21\u7d44<\/strong>
from sklearn import tree<\/p>\n\n\n\n
< model>= tree.DecisionTreeClassifier()
\u53c3\u6578\u5217\u8868
DecisionTreeClassifier(criterion=’gini’, splitter=’best’, max_depth=None, min_samples_split=2,min_samples_leaf=1, min_weight_fraction_leaf=0.0, max_features=None, random_state=None, max_leaf_nodes=None,class_weight=None, presort=False)
\u53c3\u6578\u8aaa\u660e
criterion: (default is “gini”),”gini” for the Gini impurity and “entropy” for the information gain.
\u3000gini ,\u8a08\u7b97\u5feb
\u3000entropy ,\u5207\u5272\u6548\u679c\u6bd4gini\u597d,\u4f46\u8a08\u7b97\u901f\u5ea6\u8f03\u4e45<\/p>\n\n\n\n
< model>.fit(intput, output)<\/p>\n\n\n\n
< model>.score(intput, output)<\/p>\n\n\n\n
< model>.predict(input )
\u986f\u793a\u7279\u5fb5\u91cd\u8981\u6027, \u6578\u5b57\u8d8a\u9ad8\u8868\u793a\u6b64\u6b04\u4f4d\u7279\u5fb5\u8d8a\u660e\u986f<\/strong>
print < model>.feature_importances_<\/p>\n\n\n\n
scikit-learn uses an optimised version of the CART algorithm.
http:\/\/scikit-learn.org\/stable\/modules\/generated\/sklearn.tree.DecisionTreeClassifier.html#.<\/p>\n\n\n\n\u8996\u89ba\u5316Decision Tree\u5224\u65b7\u7d50\u679c <\/strong><\/h2>\n\n\n\n
\u900f\u904e\u5716\u5f62\u986f\u793adecission tree\u7684\u6c7a\u7b56\u904e\u7a0b
\u9700\u8981export_graphviz\u548cdot\u9019\u5169\u500b\u5957\u4ef6<\/p>\n\n\n\n
on ubuntu
#sudo apt-get install graphviz python-pydot<\/p>\n\n\n\n
\u7528\u6cd5\u5982\u4e0b
tree.export_graphviz(,out_file=< outpath>,feature_names=< list of feature>)<\/p>\n\n\n\nex:
import os
from sklearn.datasets import load_iris
iris = load_iris()
from sklearn import tree
clf = tree.DecisionTreeClassifier()
model = clf.fit(iris.data, iris.target)
export_file = tree.export_graphviz( model ,out_file='tree.dot')<\/code><\/pre>\n\n\n\n
http:\/\/scikit-learn.org\/stable\/modules\/generated\/sklearn.tree.export_graphviz.html#sklearn.tree.export_graphviz<\/p>\n\n\n\n
\u6aa2\u8996dot\u683c\u5f0f\uff0c\u6216\u8f38\u51fa\u6210\u5716\u5f62\u3000
ex:
dot -Tpng tree.dot -o tree.png<\/p>\n\n\n\n
tree.dot\u683c\u5f0f\u7c21\u55ae\u8aaa\u660e\u5982\u4e0b\u3000\u3000
digraph Tree {
0 [label=”X[13] <= 416.0000ngini = 0.000270672616505nsamples = 100″, shape=”box”] ;
1 [label=”gini = 0.0000nsamples = 7387nvalue = [ 99. 0.]”, shape=”box”] ;
0 -> 1 ;
2 [label=”gini = 0.0000nsamples = 1nvalue = [ 0. 1.]”, shape=”box”] ;
0 -> 2 ;
}
[\u985e\u52250,\u985e\u52251]
\u7b2c13\u500b\u6b04\u4f4d\u5c0f\u65bc416,\u6703\u8b9399\u500b\u5c6c\u65bc\u985e\u52250,1\u500b\u5c6c\u65bc\u985e\u52251<\/p>\n\n\n\nExample in simple dataset<\/strong><\/h2>\n\n\n\n
class,packet,traffic
0,1,4
0, 2,3
1, 3 ,2
1, 4 ,1<\/p>\n\n\n\nimport numpy as np\nimport sys\nimport os\n\n##############\nfilepath=sys.argv[1]\nf = open(filepath)\ndataset = np.loadtxt(f,delimiter=',',skiprows=1)\ntarget=dataset[:,0]\ndata=dataset[:,1:]\n\nf.seek(0)\nlisthead=f.readlines()[0].strip().split(',')[1:]\n\nfrom sklearn import tree\nclf = tree.DecisionTreeClassifier()\nresult = clf.fit(data,target)\ntree.export_graphviz(result,out_file=filepath+'_tree.dot',feature_names=listhead)\n\n### list important feature ( result is few different every compute)\nimportfeature=clf.feature_importances_\ndictname=dict()\nfor name,value in zip(listhead,importfeature):\n\u3000dictname[name]=value\ndicsorted= sorted(dictname.iteritems(), key=lambda d:d[1], reverse = True)\nfor line in dicsorted:\n\u3000print line<\/code><\/pre>\n\n\n\n