- ## 随机森林、决策树MATLAB及Python完整代码实现
代码1：自编程序csdn1.m
treeNum = 10;
featureNum = 10;
dataNum =1000 ;
dataTrain=xlsread('NSL_train.xlsx');
dataTest =xlsread('NSL_test.xlsx');
y=RF(treeNum ,featureNum,dataNum ,dataTrain,dataTest);
fprintf('\n*****随机森林分类准确率为：%f***\n',y);
function [accuracy] = RF(treeNum ,featureNum,dataNum ,dataTrain,dataTest)
[dataAll,featureGrounp] = dataSet(dataTrain,treeNum,featureNum,dataNum);
RF = buildRandForest(dataAll,treeNum);
RF_prection = RFprection(RF,featureGrounp,dataTest);
accuracy = calAccuracy(dataTest,RF_prection);
end
function RF = buildRandForest(dataTrain,treeNum)
RF = [];
fprintf('*********正在训练随机森林，共%d课树**********\n',treeNum);
for a = 1: treeNum
data = dataTrain(:,:,a);
note = buildCartTree(data,0);
fprintf('++++++第%d课树训练完成\n',a);
RF = [RF,note];
fprintf('===============================\n');
end
fprintf('************随机森林训练完成！*******\n');
end
function note = buildCartTree(data,k)
k = k + 1;
[m,n] = size(data);
if m == 0
note = struct();
else
currentGini = calGiniIndex(data);
bestGini = 0;
featureNum = n - 1;
for a = 1:featureNum
feature_values = unique(data(:,a));
[m1,n1] = size(feature_values);
for b = 1:m1
[D1,D2] = splitData(data,a,feature_values(b,n1));
[m2,n2] = size(D1);
[m3,n3] = size(D2);
Gini_1 = calGiniIndex(D1);
Gini_2 = calGiniIndex(D2);
nowGini = (m2*Gini_1+m3*Gini_2)/m;
gGini = currentGini - nowGini;
if gGini > bestGini && m2>0 && m3>0
bestGini = gGini;
bestFeature = [a,feature_values(b,n1)];
rightData = D1;
leftData = D2;
end
end
end
if bestGini > 0
note = buildCartTree(rightData,k) ;
right = note;
note = buildCartTree(leftData,k) ;
left = note ;
s1 = 'bestFeature';
s2 = 'value';
s3 = 'rightBranch';
s4 = 'leftBranch';
s5 = 'leaf';
leafValue = [];
note = struct(s1,bestFeature(1,1),s2,bestFeature(1,2),s3,right,s4,left,s5,leafValue);
else
leafValue = data(1,n);
s1 = 'leaf';
note = struct(s1,leafValue);
end
end
end
function [dataAll,featureAll] = dataSet(dataTrain,treeNum,featureNum,dataNum)%数据集建立子函数
dataAll = zeros(dataNum,featureNum+1,treeNum);
featureAll = zeros(featureNum,1,treeNum);
for a = 1: treeNum
[data,feature] = chooseSample(dataTrain,featureNum,dataNum);
dataAll(:,:,a) = data;
featureAll(:,:,a) = feature';
end
end
function RF_prection_ = RFprection(RF,featureGrounp,dataTrain)
[m,n] = size(RF);
[m2,n2] = size(dataTrain);
RF_prection = [];
for a = 1:n
RF_single = RF(:,a);
feature = featureGrounp(:,:,a);
data = splitData2(dataTrain,feature);
RF_prection_single = [];
for b =1:m2
A = prection(RF_single,data(b,:));
RF_prection_single = [RF_prection_single;A];
end
RF_prection = [RF_prection,RF_prection_single];
end
RF_prection_ = mode(RF_prection,2);
end
function [Data1,Data2] = splitData(data,fea,value)
D1 = [];
D2 = [];
[m,n] = size(data);
if m == 0
D1 = 0;
D2 = 0;
else
D1 = find(data(:,fea) >= value);
D2 = find(data(:,fea) < value);
Data1 = data(D1,:);
Data2 = data(D2,:);
end
end
function data = splitData2(dataTrain,feature)
[m,n] = size(dataTrain);
[m1,n1] = size(feature);
data = zeros(m,m1);
data(:,:) = dataTrain(:,feature);
end
function [data,feature] = chooseSample(data1,featureNum,dataNum)
[m,n] = size(data1);
B = randperm(n-1);
feature = B(1,1:featureNum);
C= zeros(1,dataNum);
A = randperm(m);
C(1,:) = A(1,1:dataNum);
data= data1(C,feature);
data = [data,data1(C,n)];
end
function Gini = calGiniIndex(data)
[m,n] = size(data);
if m == 0
Gini = 0;
else
labelsNum = labels_num2(data);
[m1,n1] = size(labelsNum);
Gini = 0;
for a = 1:m1
Gini = Gini + labelsNum(a,n1)^2;
end
Gini = 1 - Gini/(m^2);
end
end
%统计标签中不同类型标签的数量
function labelsNum = labels_num2(data)
[m,n] = size(data);
if m == 0
labelsNum = 0;
else
labels = data(:,n);
A = unique(labels,'sorted');
[m1,n1] = size(A);
B = zeros(m1,2);
B(:,1) = A(:,1);
for a = 1:m1
B(a,2) = size(find(labels == A(a,1)),1);
end
labelsNum = B;
end
end
function A = prection(RF_single,sample)
if isempty(RF_single.leaf) == 0
A = RF_single.leaf;
else
B = sample(1,RF_single.bestFeature);
if B >= RF_single.value
branch = RF_single.rightBranch;
else
branch = RF_single.leftBranch;
end
A = prection(branch,sample);
end
end
function accuracy = calAccuracy(dataTest,RF_prection)
[m,n] = size(dataTest);
A = dataTest(:,n);
right = sum(A == RF_prection);
accuracy = right/m;
end
代码2：TreeBagger
%训练模型
Factor = TreeBagger(treeNumber, train, trainLabel,'Method','classification','NumPredictorsToSample',featureNum,'OOBpredictorImportance','on');%
%性能评估，k-fold交叉验证法
[Predict_label,Scores] = predict(Factor, test);%%%%测试集预测标签
代码3：Sklearn
import numpy as np
import pandas as pd
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
from sklearn.model_selection import KFold
from sklearn.metrics import roc_auc_score
from sklearn import metrics # 分类结果评价函数
#数据集读取、交叉验证等省略
forest_1 = RandomForestClassifier(n_estimators=2000, random_state=10, n_jobs=-1, oob_score=True)
forest_1.fit(x_train, y_train)
expected = y_test # 测试样本的期望输出
predicted = forest_1.predict(x_selected_test_3) # 测试样本预测
# 输出结果
print(metrics.classification_report(expected, predicted)) # 输出结果，精确度、召回率、f-1分数
print(metrics.confusion_matrix(expected, predicted)) # 混淆矩阵
auc = metrics.roc_auc_score(y_test, predicted)
accuracy = metrics.accuracy_score(y_test, predicted) # 求精度
print("RF_Accuracy: %.2f%%" % (accuracy * 100.0))
