std::vector<std::string> NaiveBayes::eval()
{
std::vector<std::string> results;
for (Data::iterator rowIt = testSet->begin(); rowIt != testSet->end() ; rowIt++)
{
DataRow &row = *rowIt;
StrVector labels = this->model->getLabels();
std::string maxLabel = labels[0];
float maxProb = 0.0f;
for (StrVector::iterator labelIt = labels.begin(); labelIt != labels.end(); labelIt++)
{
float prob = this->model->getProbability(&row, *labelIt);
if (prob > maxProb)
{
maxProb = prob;
maxLabel = *labelIt;
}
}
results.push_back(maxLabel);
}
return results;
}
void reportNGSAnalysis(const char *file_name, Params ¶ms, NGSAlignment &aln, NGSTree &tree,
DoubleMatrix &rate_info, StrVector &rate_name) {
ofstream out(file_name);
out.setf(ios::fixed,ios::floatfield);
int i, j, k;
double *rate_param = new double[aln.num_states * aln.num_states];
double *rate_matrix = new double[aln.num_states * aln.num_states];
out << "Input file: " << params.ngs_file << endl;
out << "Model of evolution: " << tree.getModel()->name << endl << endl;
out << "Substitution process assuming one homogeneous model among all positions:" << endl;
out << "Rate parameters: " << endl;
tree.getModel()->getRateMatrix(rate_param);
if (tree.getModel()->name == "UNREST") {
for (i = 0, k=0; i < aln.num_states; i++)
for (j = 0; j < aln.num_states; j++)
if (i != j)
rate_matrix[i*aln.num_states+j] = rate_param[k++];
} else {
for (i = 0, k=0; i < aln.num_states-1; i++)
for (j = i+1; j < aln.num_states; j++, k++)
rate_matrix[i*aln.num_states+j] = rate_matrix[j*aln.num_states+i] = rate_param[k];
}
for (i = 0; i < aln.num_states; i++) {
for (j = 0; j < aln.num_states; j++) {
if (j > 0) out << " \t";
if (j != i) out << rate_matrix[i*aln.num_states+j];
else out << "-";
}
out << endl;
}
out << endl;
out << "State frequencies: ";
switch (tree.getModel()->getFreqType()) {
case FREQ_EMPIRICAL:
out << "(empirical counts from alignment)" << endl;
break;
case FREQ_ESTIMATE:
out << "(estimated with maximum likelihood)" << endl;
break;
case FREQ_USER_DEFINED:
out << "(user-defined)" << endl;
break;
case FREQ_EQUAL:
out << "(equal frequencies)" << endl;
break;
default:
break;
}
double *state_freq = new double[aln.num_states];
tree.getModel()->getStateFrequency(state_freq);
for (i = 0; i < aln.num_states; i++) out << state_freq[i] << " \t";
out << endl << endl;
out << "Q matrix can be obtained by multiplying rate parameters with state frequencies" << endl << endl;
double *q_mat = new double[tree.aln->num_states * tree.aln->num_states];
tree.getModel()->getQMatrix(q_mat);
for (i = 0, k = 0; i < tree.aln->num_states; i++) {
for (j = 0; j < tree.aln->num_states; j++, k++)
out << " " << q_mat[k];
out << endl;
}
delete [] q_mat;
out << endl;
out << "Log-likelihood: " << tree.computeLikelihood() << endl << endl;
out << "Inferred posisiton-specific rates under one model or position-specific model: " << endl;
out << "Position\tSeq_error";
for (StrVector::iterator it = rate_name.begin(); it != rate_name.end(); it++)
out << "\t" << (*it);
out << endl;
for (i = 0; i < aln.ncategory; i++) {
out << i+1 << '\t' << tree.getRate()->getRate(i);
DoubleVector *rate_vec = &rate_info[i];
for (DoubleVector::iterator dit = rate_vec->begin(); dit != rate_vec->end(); dit++)
out << "\t" << *dit;
out << endl;
}
out.close();
cout << endl << "Results written to: " << file_name << endl << endl;
delete [] state_freq;
delete [] rate_matrix;
delete [] rate_param;
}
