/*
This example uses a Generalized Linear Model (GLM), in this case a LOGIT to fit to the Wisconsin Breast Cancer
data set. The output is shown below. The solution coefficients are shown.
Actual input count: 9
Actual output count: 1
Iteration #1, Error: inf
Iteration #2, Error: 0.63698368
Iteration #3, Error: 0.37629115
Iteration #4, Error: 0.24127030
Iteration #5, Error: 0.12430888
Iteration #6, Error: 0.03587100
Iteration #7, Error: 0.00312089
Ideal: 0.00, Actual: 0.02
Ideal: 0.00, Actual: 0.91
Ideal: 0.00, Actual: 0.01
Ideal: 0.00, Actual: 0.77
Ideal: 0.00, Actual: 0.02
Ideal: 1.00, Actual: 1.00
Ideal: 0.00, Actual: 0.09
Ideal: 0.00, Actual: 0.01
Ideal: 0.00, Actual: 0.01
...
Ideal: 0.00, Actual: 0.00
Ideal: 1.00, Actual: 0.94
Ideal: 0.00, Actual: 0.00
Ideal: 0.00, Actual: 0.01
Ideal: 0.00, Actual: 0.00
Ideal: 0.00, Actual: 0.00
Ideal: 1.00, Actual: 0.99
Ideal: 1.00, Actual: 0.95
Ideal: 1.00, Actual: 0.98
*/
void ExampleGLM(int argIndex, int argc, char **argv) {
char filename[FILENAME_MAX];
NORM_DATA *norm;
DATA_SET *data;
REGRESSION_MODEL *reg;
double *ideal, actual, *input, error;
unsigned int i, iteration;
LocateFile("breast-cancer-wisconsin.csv",filename,FILENAME_MAX);
/* Normalize the data. If you use a different data set, this area will need to be
updated. */
norm = NormCreate();
NormDefIgnore(norm);
NormDefPass(norm);
NormDefPass(norm);
NormDefPass(norm);
NormDefPass(norm);
NormDefPass(norm);
NormDefPass(norm);
NormDefPass(norm);
NormDefPass(norm);
NormDefPass(norm);
NormDefReplace(norm,4,1,0);
NormAnalyze(norm,filename);
data = NormProcess(norm,filename,9,1);
printf("Actual input count: %i\n", data->inputCount);
printf("Actual output count: %i\n", data->idealCount);
reg = RegressionCreate(data->inputCount,LinkLOGIT);
iteration = 0;
do {
iteration++;
error = RegressionReweightLeastSquares(reg,data);
printf("Iteration #%i, Error: %.8f\n",iteration,error);
} while (iteration < 1000 && error > 0.01);
/* Display results */
for(i=0;i<data->recordCount;i++) {
ideal = DataGetIdeal(data,i);
input = DataGetInput(data,i);
actual = RegressionCalculate(reg,input);
printf("Ideal: %.2f, Actual: %.2f\n",ideal[0],actual);
}
NormDelete(norm);
DataDelete(data);
RegressionDelete(reg);
}
/* Create the training set, and normalize the Iris data. If you are using a data set
other than Iris, you will need to update the normalization. */
static DATA_SET *create_iris_training(NORM_DATA *norm) {
char filename[FILENAME_MAX];
DATA_SET *data;
LocateFile("iris.csv",filename,FILENAME_MAX);
NormDefRange(norm,0,1);
NormDefRange(norm,0,1);
NormDefRange(norm,0,1);
NormDefRange(norm,0,1);
NormDefClass(norm,NORM_CLASS_ONEOFN,0,1);
NormAnalyze(norm,filename);
data = NormProcess(norm,filename,4,1);
return data;
}
void ExampleAnalyze(int argIndex, int argc, char **argv) {
char filename[FILENAME_MAX];
NORM_DATA *norm;
NORM_DATA_ITEM *col;
NORM_DATA_CLASS *currentClass;
LocateFile("iris.csv",filename,FILENAME_MAX);
norm = NormCreate();
NormDefRange(norm,0,1);
NormDefRange(norm,0,1);
NormDefRange(norm,0,1);
NormDefRange(norm,0,1);
NormDefClass(norm,NORM_CLASS_ONEOFN,0,1);
NormAnalyze(norm,filename);
col = norm->firstItem;
while(col!=NULL) {
if( col->type == NORM_TYPE_RANGE ) {
printf("Column: \"%s\",actualMin=%.2f,actualHigh=%.2f\n",col->name,col->actualHigh,col->actualLow);
} else {
printf("Column: \"%s\",classes= ",col->name);
currentClass = col->firstClass;
while(currentClass!=NULL) {
printf("\"%s\";",currentClass->name);
currentClass = currentClass->next;
}
printf("\n");
}
col = col->next;
}
printf("Rows: %i\n",norm->rowCount);
NormDelete(norm);
}
