%% 基於隨機森林思想的組合分類器設計 %% 清空環境變量 close all; clear; clc; %% 導入數據 Data=load('E:\study\研究生\實驗\dataset\new_housing.txt'); Label=load('E:\study\研究生\實驗\dataset\new_housingLabel.txt'); sum_Acc=0; sum_MCC=0; sum_F_measure=0; sum_G_mean=0; sum_AUC=0; [M,N]=size(Data);%數據集為一個M*N的矩陣,其中每一行代表一個樣本 indices=crossvalind('Kfold',M,5);%進行隨機分包 for k=1:5 %交叉驗證k=5,5個包輪流作為測試集 test = (indices == k); %獲得test集元素在數據集中對應的單元編號 train = ~test;%train集元素的編號為非test元素的編號 train_data=Data(train,:);%從數據集中划分出train樣本的數據 train_target=Label(train,:);%獲得樣本集的測試目標,在本例中是實際分類情況 test_data=Data(test,:);%test樣本集 test_target=Label(test,:); %模型與預測結果 %% 創建隨機森林分類器 model = classRF_train(train_data,train_target); %% 仿真測試 [Predict_label,votes] = classRF_predict(test_data,model); %預測結果概率輸出 prob_estimates=votes/500;%500為決策樹數目 output=prob_estimates(:,2);%預測為正類的概率 %調整閾值進行預測和混淆矩陣的計算 T=0.5; max_MCC=0;%記錄最大的MCC值 evaluation=[0,0,0,0];%四個評估指標存放地,初始化為全為0 for T=0.1:0.01:0.9%步長0.01 TP=0; FN=0; FP=0; TN=0; [r,~]=size(test_target); for i=1:r%樣本個數 if test_target(i,1)==1&&prob_estimates(i,2)>=T%本為正類,大於等於T則預測為正類 %正類的預測概率在prob_estimates第2列 TP=TP+1; elseif test_target(i,1)==1&&prob_estimates(i,2)<T%本為正類,小於T則預測為負類 FN=FN+1; elseif test_target(i,1)==-1&&prob_estimates(i,2)>=T%本為負類,大於等於T則預測為正類 FP=FP+1; else %即tsetLabel(i,1)==-1&&prob_estimates(i,1)<T%本為負類小於T則預測為負類 TN=TN+1; end end TP FN FP TN %Sen=TP/(TP+FN); %Spe=TN/(TN+FP); MCC=(TP*TN-FP*FN)/sqrt((TP+FP)*(TP+FN)*(TN+FP)*(TN+FN)) if MCC>max_MCC%選擇MCC值最大的那一組評估指標值 max_MCC=MCC; Precision=TP/(TP+FP); Recall=TP/(TP+FN); TPR=TP/(TP+FN); TNR=TN/(TN+FP); Acc=(TP+TN)/(TP+TN+FP+FN); F_measure=(2*Precision*Recall)/(Precision+Recall); G_mean=sqrt(TPR*TNR); evaluation(1,1)=max_MCC; evaluation(1,2)=Acc; evaluation(1,3)=F_measure; evaluation(1,4)= G_mean; end end auc=AUC(test_target,output);%每一次分類結束后進行一次計算 sum_AUC=sum_AUC+auc; sum_MCC=sum_MCC+ evaluation(1,1); sum_Acc=sum_Acc+ evaluation(1,2); sum_F_measure=sum_F_measure+ evaluation(1,3); sum_G_mean=sum_G_mean+evaluation(1,4); end avg_Acc=sum_Acc/5 avg_MCC= sum_MCC/5 avg_F_measure=sum_F_measure/5 avg_G_mean=sum_G_mean/5 avg_AUC=sum_AUC/5