matlab-基于模糊控制和决策树的预测算法matlab仿真-源码

clc; clear; close all; warning off; addpath(genpath(pwd)); B = load('dat.csv'); B=B(:,2); xx=zeros(3000,12); for i=1:3000 for j=1:12 xx(i,j)=B(i+j); end; end; [N,~]=size(xx); ntrain=round(N-200); mm=20; for i=1:ntrain for j=1:3 x11(i,j)=xx(i,j); x12(i,j)=xx(i,j+1); x13(i,j)=xx(i,j+2); x14(i,j)=xx(i,j+3); x15(i,j)=xx(i,j+4); x16(i,j)=xx(i,j+5); x17(i,j)=xx(i,j+6); x18(i,j)=xx(i,j+7); x19(i,j)=xx(i,j+8); end; y(i,1)=xx(i,12); end; [zb11 rg11]=wmdeepzb(mm,x11,y); [zb12 rg12]=wmdeepzb(mm,x12,y); [zb13 rg13]=wmdeepzb(mm,x13,y); [zb14 rg14]=wmdeepzb(mm,x14,y); [zb15 rg15]=wmdeepzb(mm,x15,y); [zb16 rg16]=wmdeepzb(mm,x16,y); [zb17 rg17]=wmdeepzb(mm,x17,y); [zb18 rg18]=wmdeepzb(mm,x18,y); [zb19 rg19]=wmdeepzb(mm,x19,y); x21(1:ntrain,1)=wmdeepyy(mm,zb11,rg11,x11); x21(1:ntrain,2)=wmdeepyy(mm,zb12,rg12,x12); x21(1:ntrain,3)=wmdeepyy(mm,zb13,rg13,x13); x22(1:ntrain,1)=x21(:,2); x22(1:ntrain,2)=x21(:,3); x22(1:ntrain,3)=wmdeepyy(mm,zb14,rg14,x14); x23(1:ntrain,1)=x21(:,3); x23(1:ntrain,2)=x22(:,3); x23(1:ntrain,3)=wmdeepyy(mm,zb15,rg15,x15); x24(1:ntrain,1)=x22(:,3); x24(1:ntrain,2)=x23(:,3); x24(1:ntrain,3)=wmdeepyy(mm,zb16,rg16,x16); x25(1:ntrain,1)=x23(:,3); x25(1:ntrain,2)=x24(:,3); x25(1:ntrain,3)=wmdeepyy(mm,zb17,rg17,x17); x26(1:ntrain,1)=x24(:,3); x26(1:ntrain,2)=x25(:,3); x26(1:ntrain,3)=wmdeepyy(mm,zb18,rg18,x18); x27(1:ntrain,1)=x25(:,3); x27(1:ntrain,2)=x26(:,3); x27(1:ntrain,3)=wmdeepyy(mm,zb19,rg19,x19); [zb21 rg21]=wmdeepzb(mm,x21,y); [zb22 rg22]=wmdeepzb(mm,x22,y); [zb23 rg23]=wmdeepzb(mm,x23,y); [zb24 rg24]=wmdeepzb(mm,x24,y); [zb25 rg25]=wmdeepzb(mm,x25,y); [zb26 rg26]=wmdeepzb(mm,x26,y); [zb27 rg27]=wmdeepzb(mm,x27,y); x31(1:ntrain,1)=wmdeepyy(mm,zb21,rg21,x21); x31(1:ntrain,2)=wmdeepyy(mm,zb22,rg22,x22); x31(1:ntrain,3)=wmdeepyy(mm,zb23,rg23,x23); x32(1:ntrain,1)=x31(:,2); x32(1:ntrain,2)=x31(:,3); x32(1:ntrain,3)=wmdeepyy(mm,zb24,rg24,x24); x33(1:ntrain,1)=x31(:,3); x33(1:ntrain,2)=x32(:,3); x33(1:ntrain,3)=wmdeepyy(mm,zb25,rg25,x25); x34(1:ntrain,1)=x32(:,3); x34(1:ntrain,2)=x33(:,3); x34(1:ntrain,3)=wmdeepyy(mm,zb26,rg26,x26); x35(1:ntrain,1)=x33(:,3); x35(1:ntrain,2)=x34(:,3); x35(1:ntrain,3)=wmdeepyy(mm,zb27,rg27,x27); [zb31 rg31]=wmdeepzb(mm,x31,y); [zb32 rg32]=wmdeepzb(mm,x32,y); [zb33 rg33]=wmdeepzb(mm,x33,y); [zb34 rg34]=wmdeepzb(mm,x34,y); [zb35 rg35]=wmdeepzb(mm,x35,y); x41(1:ntrain,1)=wmdeepyy(mm,zb31,rg31,x31); x41(1:ntrain,2)=wmdeepyy(mm,zb32,rg32,x32); x41(1:ntrain,3)=wmdeepyy(mm,zb33,rg33,x33); x42(1:ntrain,1)=x41(:,2); x42(1:ntrain,2)=x41(:,3); x42(1:ntrain,3)=wmdeepyy(mm,zb34,rg34,x34); x43(1:ntrain,1)=x41(:,3); x43(1:ntrain,2)=x42(:,3); x43(1:ntrain,3)=wmdeepyy(mm,zb35,rg35,x35); [zb41 rg41]=wmdeepzb(mm,x41,y); [zb42 rg42]=wmdeepzb(mm,x42,y); [zb43 rg43]=wmdeepzb(mm,x43,y); x51(1:ntrain,1)=wmdeepyy(mm,zb41,rg41,x41); x51(1:ntrain,2)=wmdeepyy(mm,zb42,rg42,x42); x51(1:ntrain,3)=wmdeepyy(mm,zb43,rg43,x43); [zb51 rg51]=wmdeepzb(mm,x51,y); for i=1:N for j=1:3 x11(i,j)=xx(i,j); x12(i,j)=xx(i,j+1); x13(i,j)=xx(i,j+2); x14(i,j)=xx(i,j+3); x15(i,j)=xx(i,j+4); x16(i,j)=xx(i,j+5); x17(i,j)=xx(i,j+6); x18(i,j)=xx(i,j+7); x19(i,j)=xx(i,j+8); end; y(i,1)=xx(i,12); end; x21(1:N,1)=wmdeepyy(mm,zb11,rg11,x11); x21(1:N,2)=wmdeepyy(mm,zb12,rg12,x12); x21(1:N,3)=wmdeepyy(mm,zb13,rg13,x13); x22(1:N,1)=x21(:,2); x22(1:N,2)=x21(:,3); x22(1:N,3)=wmdeepyy(mm,zb14,rg14,x14); x23(1:N,1)=x21(:,3); x23(1:N,2)=x22(:,3); x23(1:N,3)=wmdeepyy(mm,zb15,rg15,x15); x24(1:N,1)=x22(:,3); x24(1:N,2)=x23(:,3); x24(1:N,3)=wmdeepyy(mm,zb16,rg16,x16); x25(1:N,1)=x23(:,3); x25(1:N,2)=x24(:,3); x25(1:N,3)=wmdeepyy(mm,zb17,rg17,x17); x26(1:N,1)=x24(:,3); x26(1:N,2)=x25(:,3); x26(1:N,3)=wmdeepyy(mm,zb18,rg18,x18); x27(1:N,1)=x25(:,3); x27(1:N,2)=x26(:,3); x27(1:N,3)=wmdeepyy(mm,zb19,rg19,x19); x31(1:N,1)=wmdeepyy(mm,zb21,rg21,x21); x31(1:N,2)=wmdeepyy(mm,zb22,rg22,x22); x31(1:N,3)=wmdeepyy(mm,zb23,rg23,x23); x32(1:N,1)=x31(:,2); x32(1:N,2)=x31(:,3); x32(1:N,3)=wmdeepyy(mm,zb24,rg24,x24); x33(1:N,1)=x31(:,3); x33(1:N,2)=x32(:,3); x33(1:N,3)=wmdeepyy(mm,zb25,rg25,x25); x34(1:N,1)=x32(:,3); x34(1:N,2)=x33(:,3); x34(1:N,3)=wmdeepyy(mm,zb26,rg26,x26); x35(1:N,1)=x33(:,3); x35(1:N,2)=x34(:,3); x35(1:N,3)=wmdeepyy(mm,zb27,rg27,x27); x41(1:N,1)=wmdeepyy(mm,zb31,rg31,x31); x41(1:N,2)=wmdeepyy(mm,zb32,rg32,x32); x41(1:N,3)=wmdeepyy(mm,zb33,rg33,x33); x42(1:N,1)=x41(:,2); x42(1:N,2)=x41(:,3); x42(1:N,3)=wmdeepyy(mm,zb34,rg34,x34); x43(1:N,1)=x41(:,3); x43(1:N,2)=x42(:,3); x43(1:N,3)=wmdeepyy(mm,zb35,rg35,x35); x51(1:N,1)=wmdeepyy(mm,zb41,rg41,x41); x51(1:N,2)=wmdeepyy(mm,zb42,rg42,x42); x51(1:N,3)=wmdeepyy(mm,zb43,rg43,x43); yy=wmdeepyy(mm,zb51,rg51,x51); yy=yy' yy=int16(yy) plot(yy(ntrain:N),'b') hold on plot(B(ntrain+13:N+60),'r') clear all function [zb, ranges]=wmdeepzb(mm,xx,y) % Train fuzzy system with input xx and output y to get zb and ranges, % where zb is the in FS (7) and ranges is the endpoints in (11). extra=[xx,y]; [numSamples,m]=size(extra); numInput=m-1; for i=1:numInput fnCounts(i)=mm; end; ranges = zeros(numInput,2); activFns = zeros(numInput,2); activGrades = zeros(numInput,2); searchPath = zeros(numInput,2); numCells = 1; % number of regions (cells) for i = 1:numInput a=min(extra(:,2)); b=max(extra(:,2)); ranges(i,1) = min(extra(:,i)); ranges(i,2) = max(extra(:,i)); numCells = numCells * fnCounts(i); end; baseCount(1)=1; for i=2:numInput baseCount(i)=1; for j=2:i baseCount(i)=baseCount(i)*fnCounts(numInput-j+2); end; end; % Generate rules for cells covered by data zb = zeros(1,numCells); % THEN part centers of generated rules ym = zeros(1,numCells); for k = 1:numSamples for i = 1:numInput numFns = fnCounts(i);%the number of fuzzy sets nthActive = 1;%which one to activte for nthFn = 1:numFns grade = meb2(numFns,nthFn,extra(k,i),ranges(i,1),ranges(i,2)); if grade > 0 activFns(i,nthActive) = nthFn;% which fuzzy set own weights [1,20] activGrades(i,nthActive) = grade; nthActive = nthActive + 1; end; end; % endfor nthFn end; % endfor i for i=1:numInput if activGrades(i,1) >= activGrades(i,2)%select the biggest weight searchPath(i,1)=activFns(i,1);%the ith input variable searchPath(i,2)=activGrades(i,1); else searchPath(i,1)=activFns(i,2); searchPath(i,2)=activGrades(i,2); end; end; indexcell=1; grade=1; for i=1:numInput grade=grade*searchPath(i,2);%searchPath(i,2) save the activGrades indexcell=indexcell+(searchPath(numInput-i+1,1)-1)*baseCount(i);%计算集合在空间中的具体位置 end; ym(indexcell)=ym(indexcell)+grade;%线，面，片 zb(indexcell)=zb(indexcell)+extra(k,numInput+1)*grade;%用y来更新权重 end; % endfor k for j=1:numCells if ym(j) ~= 0 zb(j)=zb(j)/ym(j); end; end; % Extrapolate the rules to all the cells for i=1:numInput-1 baseCount(i)=1; for j=1:i baseCount(i)=baseCount(i)*fnCounts(numInput-j+1); end; end; ct=1; zbb = zeros(1,numCells); ymm = zeros(1,numCells); while ct > 0 ct=0; for s=1:numCells if ym(s) == 0 ct=ct+1; end; end; for s=1:numCells if ym(s) == 0 s1=s; index = ones(1,numInput); for i=numInput-1:-1:1 while s1 > baseCount(i) s1=s1-baseCount(i); index(numInput-i)=index(numInput-i)+1; end; end; index(numI