Применение нейронных сетей к оценке кредитоспособности физических лиц

{-4,67149656505120e-001, -6,09590774067550e+000, 1,75210100895654e+000, -3,13956250775567e+000, 1,43068245551182e+001, -7,85053944357368e-001, 4,24516927798281e+000, -2,59941858924333e+000, 6,03602215289459e-001, -2,18964575431617e+000, 1,14808888593193e+001, 7,03883507225023e+000 },

{-1,00981535367063e+000, -6,32535934729304e+000, 4,01064987145387e+000, 1,48465851842388e+000, 9,81770015310024e+000, 3,56768950228247e+000, 2,09026878326488e+000, -3,56947927444606e+000, 3,78673354205937e+000, 6,97345367206553e+000, 3,56882993548831e+000, 4,76152715633977e-001 },

{-8,54838370143648e-001, -6,96381955453000e+000, 3,30845287107656e+000, 8,45395665670600e-001, 1,02274795138672e+001, 2,97825904668851e+000, 2,34064744676174e+000, -4,02916055317983e+000, 3,23672074326387e+000, 5,90548752061914e+000, 4,22109537118753e+000, 7,78710016942739e-001 }

};

double hidden_bias[5]={ -4,89949635603894e-001, -1,92268683845077e+000, 4,34461990732500e+000, 2,67856825427872e+000, 2,22697833237394e+000 };

double hidden_output_wts[3][5]=

{

{-3,31754742496978e+000, 8,98228051391912e+000, 4,09646018637776e+000, -2,98424512535983e+000, -6,24495119880673e+000},

{1,29127644257301e+000, -3,38886487369035e+000, 4,27371800423730e+000, -7,54734238002245e-002, 5,79364818806007e-001 },

{2,10323923959187e+000, -5,53525376784362e+000, -8,23942580395382e+000, 3,02222658249428e+000, 5,59800454230386e+000 }

};

double output_bias[3]={ 2,55329441003075e+000, 9,52241585555783e-001, -3,56191361399268e+000 };

double max_input[12]={ 3,00000000000000e+000, 4,00000000000000e+000, 3,00000000000000e+000, 4,00000000000000e+000, 3,00000000000000e+000, 4,00000000000000e+000, 4,00000000000000e+000, 3,00000000000000e+000, 3,00000000000000e+000, 3,00000000000000e+000, 3,00000000000000e+000, 6,00000000000000e+000 };

double min_input[12]={ 1,00000000000000e+000, 1,00000000000000e+000, 0,00000000000000e+000, 1,00000000000000e+000, 1,00000000000000e+000, 1,00000000000000e+000, 1,00000000000000e+000, 1,00000000000000e+000, 1,00000000000000e+000, 1,00000000000000e+000, 1,00000000000000e+000, 1,00000000000000e+000 };

double input[12];

double hidden[5];

double output[3];

void FindMax(double* vec, double* max, long* maxIndex,int len)

{

long i;

*max = vec[0];

*maxIndex = 0;

for(i=1; i<len; i++)

{

if(vec[i]>*max)

{

*max = vec[i];

*maxIndex = i;

}

}

}

void ScaleInputs(double* input, double minimum, double maximum, int size)

{

double delta;

long i;

for(i=0; i<size; i++)

{

delta = (maximum-minimum)/(max_input[i]-min_input[i]);

input[i] = minimum - delta*min_input[i]+ delta*input[i];

}

}

void softmax(double* vec,int len)

{

long i, j;

double sum=0.0;

for(i=0; i<len; i++)

{

if(vec[i]>200)

{

double max;

long maxIndex;

FindMax(vec, &max, &maxIndex,len);

for(j=0; j<len; j++)

{ if(j==maxIndex) vec[j] = 1.0;

else vec[j] = 0.0;

}

return;

}

else

{

vec[i] = exp(vec[i]);

}

sum += vec[i];

}

if(sum==0)

{

long a = 1;

}

if(sum!=0.0)

{

for(i=0; i<len; i++) vec[i] = vec[i]/sum;

}

else for(i=0; i<len; i++) vec[i] = 1.0/(double)len;

}

double logistic(double x)

{

if(x > 100.0) x = 1.0;

else if (x < -100.0) x = 0.0;

else x = 1.0/(1.0+exp(-x));

return x;

}

void ComputeFeedForwardSignals(double* MAT_INOUT,double* V_IN,double* V_OUT, double* V_BIAS,int size1,int size2,int layer)

{

int row,col;

for(row=0;row < size2; row++)

{

V_OUT[row]=0.0;

for(col=0;col<size1;col++)V_OUT[row]+=(*(MAT_INOUT+(row*size1)+col)*V_IN[col]);

V_OUT[row]+=V_BIAS[row];

if(layer==0) V_OUT[row] = logistic(V_OUT[row]);

}

}

void RunNeuralNet_Classification ()

{

ComputeFeedForwardSignals((double*)input_hidden_weights,input,hidden,hidden_bias,12, 5,0);

ComputeFeedForwardSignals((double*)hidden_output_wts,hidden,output,output_bias,5, 3,1);

}

int main()

{

int index;

int i=0;

int keyin=1;

double max;

while(1)

{

max=3.e-300;

printf("\nEnter values for Continuous inputs\n");

printf("Cont. Input-0(Var1): ");

scanf("%lg",&input[0]);

printf("Cont. Input-1(Var2): ");

scanf("%lg",&input[1]);

printf("Cont. Input-2(Var3): ");

scanf("%lg",&input[2]);

printf("Cont. Input-3(Var4): ");

scanf("%lg",&input[3]);

printf("Cont. Input-4(Var5): ");

scanf("%lg",&input[4]);

printf("Cont. Input-5(Var6): ");

scanf("%lg",&input[5]);

printf("Cont. Input-6(Var7): ");

scanf("%lg",&input[6]);

printf("Cont. Input-7(Var8): ");

scanf("%lg",&input[7]);

printf("Cont. Input-8(Var9): ");

scanf("%lg",&input[8]);

printf("Cont. Input-9(Var10): ");

scanf("%lg",&input[9]);

printf("Cont. Input-10(Var11): ");

scanf("%lg",&input[10]);

printf("Cont. Input-11(Var12): ");

scanf("%lg",&input[11]);

ScaleInputs(input,0,1,12);

RunNeuralNet_Classification();

//Output Activation is Softmax;

softmax(output, 3);

for(i=0;i<3;i++)

{

if(max<output[i])

{

max=output[i];

index=i+1;

}

}

printf("\nPredicted category = ");

switch(index)

{

case 1: printf("A\n"); break;

case 2: printf("B\n"); break;

case 3: printf("C\n"); break;

default: break;

}

printf("\nConfidence level = %.14f",max);

printf("\n\nPress any key to make another prediction or enter 0 to quit the program.\n");

keyin=getch();

if(keyin==48)break;

}

return 0;

}

Размещено на Allbest.ru