三支決策算法(ThreeDecisonWay)用于多分類任務(wù)的Matlab程序

1、三支決策算法(ThreeDecisonWay)用于多分類任務(wù)的Matlab程序functionresult=TDW_multiclass(TrainingData_File)%ThisisafunctionexpandTDWtomulticlass%該函數(shù)是基于one-vs-one方法的得到的處理多類問題的TDW分類器%trainX訓(xùn)練樣本的特征向量構(gòu)成的n行m列矩陣，每一行是一個(gè)樣本%trainY訓(xùn)練樣本的標(biāo)簽構(gòu)成的n行Q列矩陣，每一行對(duì)應(yīng)一個(gè)樣本，Q表示類別的個(gè)數(shù),%如果第i個(gè)樣本屬于第j類則trainY(i,j)=1,且trainY(i,:)中的其它元素都為-1%Getthedetail

2、ofdatasettrain_data=load(TrainingData_File);Y,=Data(train_data);trainY=Y;trainX=X;n,m=size(trainY);Sample_size=zeros(m,1);Class=,;Sample_area=zeros(n,m);%Gettheareaof2classofsamplesforr=1:m,flagp=;flagpY=;fori=1:niftrainY(i,r)=1flagp=flagp;trainX(i,:);flagpY=flagpY;trainY(i,:);endendnp=size(flagp,1)

3、;ifr+1Rho&Ratio_BN10*Rho&Ratio_BN1result=Overlapping;elseresult=Inter|Intar;enddisp(DatasetCategory=,result);endfunctionDatasetCategory=TDW_Func(TrainingData_File)%*Thisfunctionaimtoclassifydataset(binaryonly)*%Setdatasetandinitialization%TrainingData_File=adult.csv;train_data=load(TrainingData_File

4、);train_target,P,NumberofData,NumberofInputNeurons,=Data(train_data);n=0;k=0;%DistancebetweeneachsampleandtherestofthesampleDistance=zeros(NumberofData-1,NumberofData);fori=1:1:NumberofDataforj=1:1:NumberofDatasum=0;ifijfork=1:NumberofInputNeuronssum=sum+power(P(k,i)-P(k,j),2);Distance(j,i)=sqrt(sum

5、);endelseifijfork=1:NumberofInputNeuronssum=sum+power(P(k,i)-P(k,j),2);Distance(j-1,i)=sqrt(sum);endendendend%size(Distance)%Distance1=Distance;%Distance(Distance=0)=;%size(Distance)%Distance=reshape(Distance,NumberofData-1,NumberofData);%Thei-thcolumnisthedistance%ofthei-1thsampleandtheremainingi-1

6、samples%Determiningthevalueofneighborhood%Distance_sorted=zeros(NumberofData-1.NumberofData);w=0.1;%Range(0,1),itthekeytodeterminingthenumberofneigborhoodsamplesw=0.1;%Range(0,1),itthekeytodeterminingthenumberofneigborhoodsamplesfori=1:NumberofDataDistance_c=Distance(:,i);table=tabulate(Distance_c);

7、n,m=size(table);Distance_sorted(1:n,i)=table(:,1);%ArrangethrdistancefromlargetosmallendDelata=zeros(1,NumberofData);%ForeverysamplehasaDelatafori=1:NumberofDataDelata(1,i)=min(Distance_sorted(:,i)+w*(max(Distance_sorted(:,i)-min(Distance_sorted(:,i);end%GetthesamplebelongstotheneighborhoodDistance_

8、neig=zeros(NumberofData-1,NumberofData,2);fori=1:NumberofDatak=1;forj=1:NumberofData-1ifDistance(j,i)Delata(1,i)Distance_neig(k,i,1)=Distance(j,i);ifjiDistance_neig(k,i,2)=train_target(1,j);elseDistance_neig(k,i,2)=train_target(1,j+1);endk=k+1;endendend%Determiningwhichareathesampleisalph=5;beta=-5/

9、6;%PartitionparameterNumberofPos=0;NumberofBnd=0;NumberofNeg=0;%Initializethenumberofdifferentareasamplefx=zeros(NumberofData,1);table=tabulate(train_target(1,:);iftable(1,2)alphNumberofPos=NumberofPos+1;PosData(1,NumberofPos)=train_target(1,i);PosData(2:NumberofInputNeurons+1,NumberofPos)=P(:,i);el

10、seiffx(i)=betaNumberofBnd=NumberofBnd+1;NumberofBnd=NumberofBnd+1;BndData(1,NumberofBnd)=train_target(1,i);BndData(2:NumberofInputNeurons+1,NumberofBnd)=P(:,i);elseiffx(i)betaNumberofNeg=NumberofNeg+1;NegData(1,NumberofNeg)=train_target(1,i);NegData(2:NumberofInputNeurons+1,NumberofNeg)=P(:,i);enden

11、difNumberofNeg0.7*NumberofDataDatasetCategory=Inter_calss;elseifNumberofNeg0.1*NumberofData&NumberofBnd0.3*NumberofData&NumberofNegjfork=1:NumberofInputNeuronssum=sum+power(P(k,i)-P(k,j),2);Distance(j,i)=sqrt(sum);endelseifijfork=1:NumberofInputNeuronssum=sum+power(P(k,i)-P(k,j),2);Distance(j-1,i)=s

12、qrt(sum);endendendend%Determiningthevalueofneighborhood%Distance_sorted=zeros(NumberofData-1.NumberofData);w=0.05;%Range(0.01,0.05),itsthekeytodeterminingthenumberofneigborhoodsamplesfori=1:NumberofDataDistance_c=Distance(:,i);table=tabulate(Distance_c);n,m=size(table);Distance_sorted(1:n,i)=table(:

13、,1);%ArrangethedistancefromlargetosmallendDelata=zeros(1,NumberofData);%ForeverysamplehasaDelatafori=1:NumberofDataDelata(1,i)=min(Distance_sorted(:,i)+w*(max(Distance_sorted(:,i)-min(Distance_sorted(:,i);end%GetthesamplebelongstotheneighborhoodDistance_neig=zeros(NumberofData-1,NumberofData,2);fori

14、=1:NumberofDatak=1;forj=1:NumberofData-1ifDistance(j,i)Delata(1,i)Distance_neig(k,i,1)=Distance(j,i);ifjiDistance_neig(k,i,2)=train_target(1,j);elseDistance_neig(k,i,2)=train_target(1,j+1);endk=k+1;endendend%UseKNNtogettheKnearestsamplek=15;%ifuseKNN,thekisakeypartemeterIDX=knnsearch(P,P,K,k,Distanc

15、e,euclidean);%返回每個(gè)樣本的K近鄰樣本，每行代表每個(gè)樣本的K近鄰樣本的索引值%Determiningwhichareathesampleisalph=K;beta=-K/(K+1);NumberofPos=0;NumberofBnd=0;NumberofNeg=0;fx=zeros(NumberofData,1);%正域與邊界域的閾值%Partitionparameter%Initializethenumberofdifferentareasampletable=tabulate(train_target(1,:);iftable(1,2)=alph%正域判斷NumberofPos=NumberofPos+1;PosData(1,NumberofPos)=train_target(1,i);PosData(2:NumberofInputNeurons+1,NumberofPos)=P(:,i);PBN(i,1)=1;elseiffx(i)beta%邊界域判斷NumberofBnd=NumberofBnd+1;BndData(1,NumberofBnd)=train_ta