double DecisionTree::ComputeEntropyGivenSurvival(int attr, vi& data){
vi ones, zeroes;
for(int i : data){
if(train[i][0] == 0)
zeroes.pb(i);
else
ones.pb(i);
}
double entropy = (ones.size() * ComputeEntropy(attr, ones) + zeroes.size() * ComputeEntropy(attr, zeroes)) / data.size();
return entropy;
}
// LogZp
double MatchMRFGraph::ComputeLogZp()
{
double ae,en;
ae = ComputeAverageEnergy();
en = ComputeEntropy();
return ae+en;
}
int DecisionTree::ChooseAttributeToSplit(node* curr){
if(__MESSAGE__)
cout << "Choosing an attribute to split ... " << endl;
vector<pair<double,int> > v;
double mutualInfo = 0;
for(int i=1; i<NUM_ATTR; i++){
if(curr->parAttr.count(i))
continue;
mutualInfo = ComputeEntropy(i, curr->data) - ComputeEntropyGivenSurvival(i, curr->data);
v.pb(mp(mutualInfo, NUM_ATTR - i));
}
if(v.empty())
return -1;
sort(v.begin(), v.end());
int ret = (NUM_ATTR - v.back().second);
if(__MESSAGE__)
cout << "Chosen attribute number for splitting : " << ret << "..." <<endl;
return ret;
}
