Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

new AUC evaluator, now compatible with weighted loss

Browse Source
This commit is contained in:
tqchen 2014-04-29 17:03:34 -07:00
parent 31edfda03c
commit 87a9c22795
1 changed files with 42 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

64
regrank/xgboost_regrank_eval.h
View File

@ -94,34 +94,54 @@ namespace xgboost{
} }
}; };
/*! \brief Area under curve */ /*! \brief Area under curve, for both classification and rank */
struct EvalAuc : public IEvaluator{ struct EvalAuc : public IEvaluator{
inline static bool CmpFirst( const std::pair<float,float> &a, const std::pair<float,float> &b ){ inline static bool CmpFirst( const std::pair<float,unsigned> &a, const std::pair<float,unsigned> &b ){
return a.first > b.first; return a.first > b.first;
} }
virtual float Eval( const std::vector<float> &preds, virtual float Eval( const std::vector<float> &preds,
const DMatrix::Info &info ) const { const DMatrix::Info &info ) const {
const std::vector<float> &labels = info.labels; std::vector<unsigned> tgptr(2,0); tgptr[1] = preds.size();
const unsigned ndata = static_cast<unsigned>( preds.size() ); const std::vector<unsigned> &gptr = info.group_ptr.size() == 0 ? tgptr : info.group_ptr;
std::vector< std::pair<float, float> > rec; utils::Assert( gptr.back() == preds.size(), "EvalAuc: group structure must match number of prediction" );
for( unsigned i = 0; i < ndata; ++ i ){ const unsigned ngroup = static_cast<unsigned>( gptr.size() );
rec.push_back( std::make_pair( preds[i], labels[i]) );
} double sum_auc = 0.0f;
random::Shuffle( rec ); #pragma omp parallel reduction(+:sum_auc)
std::sort( rec.begin(), rec.end(), CmpFirst ); {
// each thread takes a local rec
long npos = 0, nhit = 0; std::vector< std::pair<float,unsigned> > rec;
for( unsigned i = 0; i < ndata; ++ i ){ #pragma omp for schedule(static)
if( rec[i].second > 0.5f ) { for( unsigned k = 0; k < ngroup; ++ k ){
++ npos; rec.clear();
}else{ for( unsigned j = gptr[k]; j < gptr[k+1]; ++ j ){
// this is the number of correct pairs rec.push_back( std::make_pair( preds[j], j ) );
nhit += npos; }
std::sort( rec.begin(), rec.end(), CmpFirst );
// calculate AUC
double sum_pospair = 0.0;
double sum_npos = 0.0, sum_nneg = 0.0, buf_pos = 0.0, buf_neg = 0.0;
for( size_t j = 0; j < rec.size(); ++ j ){
const float wt = info.GetWeight( rec[j].second );
const float ctr = info.labels[ rec[j].second ];
// keep bucketing predictions in same bucket
if( j != 0 && rec[j].first != rec[j-1].first ){
sum_pospair += buf_neg * (sum_npos + buf_pos *0.5);
sum_npos += buf_pos; sum_nneg += buf_neg;
buf_neg = buf_pos = 0.0f;
}
buf_pos += ctr * wt; buf_neg += (1.0f-ctr) * wt;
}
sum_pospair += buf_neg * (sum_npos + buf_pos *0.5);
sum_npos += buf_pos; sum_nneg += buf_neg;
//
utils::Assert( sum_npos > 0.0 && sum_nneg > 0.0, "the dataset only contains pos or neg samples" );
// this is the AUC
sum_auc += sum_pospair / (sum_npos*sum_nneg);
} }
} }
long nneg = ndata - npos; // return average AUC over list
utils::Assert( nneg > 0, "the dataset only contains pos samples" ); return static_cast<float>(sum_auc) / ngroup;
return static_cast<float>(nhit) / nneg / npos;
} }
virtual const char *Name( void ) const{ virtual const char *Name( void ) const{
return "auc"; return "auc";