improvedran.asv

改进RAN应用股票财务程序

2618.45
2624.41
2661.36
2663.99
2652.30
2661.84
2673.52
2655.40
2664.23
2636.81
2649.21
2612.25
2643.01
2624.89
2624.65
2615.59
2634.18
2658.27
2652.30
2649.44
2649.44
2661.60
2649.21
2658.50
2661.36
2649.21
2686.16
2686.16
2637.05
2683.06
2646.34
2679.96
2707.62
2373.35
2343.07
2360.95
2339.49
2288.23
2222.91
2250.80
2235.30
2281.80
2257.00
2241.50
2235.78
2385.75
2379.79
2391.71
2355.23
2348.79
2315.18
2364.05
2342.83
2351.89
2339.97
2312.31
2330.91
2305.88
2330.67
2327.57
2333.53
2361.19
2367.15
2376.69
2367.15
2364.29
2379.31
2394.81
2385.99
2391.95
2401.01
2416.74
2416.50
2440.82
2449.88
2465.38
2477.78
2459.42
2480.64
2471.82
2489.94
2489.94
2508.54
2539.05
2511.64
2499.00
2502.10
2550.98
2516.88
2495.42
2501.62
2452.98
2449.17
2428.43
2409.59
2415.79
2440.59
2418.89
2434.63
2465.14
2440.59
2421.99
2382.65
2394.81
2370.01
2394.81
2379.55
2361.19
2354.75
2348.79
2293.72
2305.88
2330.43
2360.95
2367.15
2366.91
2394.81
2404.11
2456.32
2428.43
2428.66
2670.66
2670.66
2624.65
2624.65
2539.05
2538.82
2551.21
2581.73
2560.04
2551.21
2502.10
2508.06
2495.90
2504.96
2514.50
2520.70
2520.70
2502.10
2514.50
2495.90
2538.82
2508.30
2501.86
2474.44
2425.56
2388.61
2413.41
2398.15
2379.79
2351.89
2343.07
2318.28
2318.28
2312.08
2391.95
2391.71
2376.69
2345.93
2349.03
2330.91
2315.18
2312.31
2290.86
2354.99
2413.41
2407.21
2391.71
2407.21
2401.01
2379.79
2343.07
2343.07
2343.07
2324.47
2330.43
2305.88
2318.28
2287.76
2303.02
2336.63
2324.71
2297.06
2318.04
2300.16
2293.96
2297.06
2284.66
2287.76
2293.96
2290.86
2293.72
2281.32
2254.14
2275.36
2281.80
2275.36
2278.94
2303.02
2263.20
2266.06
2223.62
2232.44
2257.24
2241.74
2257.24
2281.56
2260.10
2266.06
2284.42
2305.88
2308.98
2312.08
2318.04
2330.43
2305.88
2388.85
2398.15
2452.98
2440.82
2465.14
2514.26
]';
[pn,meanp,stdp,tn,meant,stdt]=prestd(p,t); %生成均值为0，方差为一的矩阵
[ptrans,transMat] = prepca(pn,0.01) ;%需调整的参数-2；主元分析
Tn=(t-min(t))/(max(t)-min(t));  
[s,l]=size(ptrans);
ptrans=(ptrans-repmat(min(ptrans')',1,l))./repmat(max(ptrans')'-min(ptrans')',1,l);  %  矩阵归一化
p1=ptrans(:,1:300);
p2=ptrans(:,301:396);
t1=Tn(:,1:300);
t2=Tn(:,301:396);
[InDim,TrainSamNum]=size(p1);
[OutDim,TrainSamNum]=size(t1);
[InDim,TestSamNum]=size(p2);
TrainSamNum;      %训练样本数
TestSamNum;       %测试样本数
InDim;            %样本输入维数
OutDim;           %样本输出维数

%根据目标函数获取样本输入输出

TestSTD=std(t2);  %  std(x),x为向量,std表示x方差的无偏估计平方根

OverLapCoe=0.8;  %重叠系数
Dist_Max=1.5;  %最大距离分辨率
Dist_Min=0.11;  %最小距离分辨率
ErrLimit=0.02;  %误差分辨率
Decay=0.977;  %分辨率衰减常数
lr=0.05;  %学习率
MaxEpoch=100;  %最大学习次数
DistLimit=Dist_Max;  %距离分辨率
b2=t1(:,1);
w2=[];
UnitCenters=[];
SpreadConstant=[]; 
UnitNum=0;
AllUnitNum=0;
AllTestRSME=[];
tp=[ErrLimit lr MaxEpoch];
for TrainedNum=2:TrainSamNum
    NewInput=p1(:,TrainedNum);
    NewOutput=t1(:,TrainedNum);
    NetOut=RBFNN(NewInput,UnitCenters,w2,b2,SpreadConstant);
    NewErr=NewOutput-NetOut;
    if (UnitNum==0),
        NewDist=Dist_Max;
    else
        AllDist=dist(UnitCenters',NewInput);
        NewDist=min(AllDist);
    end
    if(norm(NewErr)>=ErrLimit & NewDist>=DistLimit),  %判断是否添加隐节点
        [UnitCenters,w2,SpreadConstant]=AddNewUnit(NewInput,NewErr,NewDist,UnitCenters,w2,SpreadConstant,OverLapCoe);
        TrainedNum;
        UnitNum=UnitNum+1;
    else
        [UnitCenters,w2,b2]=FineTuning(NewInput,NewOutput,UnitCenters,w2,b2,SpreadConstant,tp);  %  参数精调的每一次迭代都是一个样本进入
    end
    
    if DistLimit>Dist_Min,  %分辨率衰减
        DistLimit=DistLimit*Decay;
    else
        DistLimit=Dist_Min;
    end
    AllUnitNum=[AllUnitNum UnitNum];
    
    TestNNOut=RBFNN(p2,UnitCenters,w2,b2,SpreadConstant);
    TestRSME=sqrt(sumsqr(TestNNOut-t2)/TestSamNum)/TestSTD;
    AllTestRSME=[AllTestRSME TestRSME];
end