std::vector<std::string> GreedyStepwiseSearch::findSubset(const DataFrame& df,
TgsProgress* p)
{
TgsProgress* dataFrameProgress = NULL;
TgsProgress* evaluateProgress = NULL;
if (p)
{
dataFrameProgress = p->createTgsChild("Data Frame Processing", .5);
evaluateProgress = p->createTgsChild("Evaluate Subsets", .5);
}
std::vector<std::string> result;
_fse->setDataFrame(df, dataFrameProgress);
assert(_direction == Backward);
_progress = evaluateProgress;
_iteration = 0;
_totalIterations = (df.getNumFactors() * (df.getNumFactors() + 1)) / 2;
double bestScore = -1e300;
vector<int> bestSolution;
if (_direction == Backward)
{
vector<int> columns;
for (unsigned int i = 0; i < df.getNumFactors(); i++)
{
columns.push_back(i);
}
double score = _fse->evaluateSubset(columns);
bestScore = score;
bestSolution = columns;
// printf("score: %.3f size: %d\n", score, columns.size());
do
{
score = _removeWorst(columns);
// printf("score: %.3f size: %d\n", score, columns.size());
for (unsigned int i = 0; i < columns.size(); i++)
{
string s = df.getFactorLabelFromIndex(columns[i]);
char* s2 = (char*)s.c_str();
s2[4] = 0;
// printf("%s\t", s2);
}
// printf("\n");
if (score >= bestScore)
{
bestScore = score;
bestSolution = columns;
}
} while (columns.size() > 1);
}
result.clear();
for (unsigned int i = 0; i < bestSolution.size(); i++)
{
result.push_back(df.getFactorLabelFromIndex(bestSolution[i]));
}
if (p)
{
p->setProgress(1.0);
}
return result;
}
