PyObject* UserKnn_testrec( PyUserKnn* self, PyObject* args, PyObject* kwdict )
{
const char* input_file = NULL;
const char* output_file = NULL;
char dlmchar = ',';
int header = 0;
int usercol = 0;
int itemcol = 1;
int ratingcol = -1;
int topn = 10;
static char* kwlist[] = { const_cast<char*>( "input_file" ),
const_cast<char*>( "output_file" ),
const_cast<char*>( "dlmchar" ),
const_cast<char*>( "header" ),
const_cast<char*>( "usercol" ),
const_cast<char*>( "itemcol" ),
const_cast<char*>( "ratingcol" ),
const_cast<char*>( "topn" ),
NULL };
if( !PyArg_ParseTupleAndKeywords( args, kwdict, "s|sciiiii", kwlist, &input_file,
&output_file, &dlmchar, &header, &usercol, &itemcol, &ratingcol, &topn ) )
{
return NULL;
}
if( NULL == input_file )
{
return NULL;
}
DataWriter dataWriter;
if( NULL != output_file )
{
string strfilename = output_file;
dataWriter.open( strfilename );
}
DataReader testReader( input_file, dlmchar, header );
DataFrame testData( testReader, usercol, itemcol, ratingcol );
PyObject* pyDict = PyDict_New();
if( NULL == pyDict )
{
return NULL;
}
map<string, int> userFilter;
DataFrame::iterator ind;
DataFrame::iterator end = testData.end();
for( ind = testData.begin() ; ind != end ; ++ind )
{
// Recommned item list once per user
std::string userId = ind->first.first;
if( userFilter.find( userId ) != userFilter.end() )
{
continue;
}
userFilter[userId] = 0;
vector<string> ranking;
if( !self->m_recAlgorithm->recommend( userId, topn, ranking ) )
{
continue;
}
PyObject* pyList = PyList_New( 0 );
vector<string>::iterator rankind;
vector<string>::iterator rankend = ranking.end();
for( rankind = ranking.begin() ; rankind != rankend ; ++rankind )
{
#if PY_MAJOR_VERSION >= 3
if( -1 == PyList_Append( pyList, PyBytes_FromString( rankind->c_str() ) ) )
#else
if( -1 == PyList_Append( pyList, PyString_FromString( rankind->c_str() ) ) )
#endif
{
return NULL;
}
}
if( PyDict_SetItemString( pyDict, userId.c_str(), pyList ) < 0 )
{
Py_DECREF( pyList );
return NULL;
}
if( itemcol >= 0 && ratingcol >= 0 )
{
// Calculate p@r, nDCG
;
}
if( dataWriter.isOpen() )
{
dataWriter.write( userId, ranking );
}
}
PyObject* pyTupleResult = PyTuple_New( 1 );
PyTuple_SET_ITEM( pyTupleResult, 0, pyDict );
//PyTuple_SET_ITEM( pyTupleResult, 1, PyFloat_FromDouble( self->m_mae.eval() ) );
//PyTuple_SET_ITEM( pyTupleResult, 2, PyFloat_FromDouble( self->m_rmse.eval() ) );
return pyTupleResult;
}
PyObject* UserKnn_test( PyUserKnn* self, PyObject* args, PyObject* kwdict )
{
const char* input_file = NULL;
const char* output_file = NULL;
char dlmchar = ',';
int header = 0;
int usercol = 0;
int itemcol = 1;
int ratingcol = -1;
static char* kwlist[] = { const_cast<char*>( "input_file" ),
const_cast<char*>( "output_file" ),
const_cast<char*>( "dlmchar" ),
const_cast<char*>( "header" ),
const_cast<char*>( "usercol" ),
const_cast<char*>( "itemcol" ),
const_cast<char*>( "ratingcol" ),
NULL };
if( !PyArg_ParseTupleAndKeywords( args, kwdict, "s|sciiii", kwlist, &input_file,
&output_file, &dlmchar, &header, &usercol, &itemcol, &ratingcol ) )
{
return NULL;
}
if( NULL == input_file )
{
return NULL;
}
DataWriter dataWriter;
if( NULL != output_file )
{
char dlm = '\t';
string strfilename = output_file;
if( strfilename.substr( strfilename.find_last_of( "." ) + 1 ) == "csv" )
{
dlm = ',';
}
dataWriter.open( strfilename, dlm );
}
DataReader testReader( input_file, dlmchar, header );
DataFrame testData( testReader, usercol, itemcol, ratingcol );
PyObject* pyList = PyList_New( 0 );
if( NULL == pyList )
{
return NULL;
}
DataFrame::iterator ind;
DataFrame::iterator end = testData.end();
for( ind = testData.begin() ; ind != end ; ++ind )
{
std::string userId = ind->first.first;
std::string itemId = ind->first.second;
double prediction = self->m_recAlgorithm->predict( userId, itemId );
PyObject* pyTuple = PyTuple_New( 3 );
if( NULL == pyTuple )
{
return NULL;
}
#if PY_MAJOR_VERSION >= 3
PyTuple_SET_ITEM( pyTuple, 0, PyBytes_FromString( userId.c_str() ) );
PyTuple_SET_ITEM( pyTuple, 1, PyBytes_FromString( itemId.c_str() ) );
#else
PyTuple_SET_ITEM( pyTuple, 0, PyString_FromString( userId.c_str() ) );
PyTuple_SET_ITEM( pyTuple, 1, PyString_FromString( itemId.c_str() ) );
#endif
PyTuple_SET_ITEM( pyTuple, 2, PyFloat_FromDouble( prediction ) );
if( -1 == PyList_Append( pyList, pyTuple ) )
{
return NULL;
}
if( ratingcol >= 0 )
{
double rating = ind->second;
self->m_mae.append( rating, prediction );
self->m_rmse.append( rating, prediction );
}
if( dataWriter.isOpen() )
{
vector<string> vline;
vline.push_back( userId );
vline.push_back( itemId );
std::ostringstream ss;
ss << prediction;
vline.push_back( ss.str() );
dataWriter.write( vline );
}
}
PyObject* pyTupleResult = PyTuple_New( 3 );
PyTuple_SET_ITEM( pyTupleResult, 0, pyList );
PyTuple_SET_ITEM( pyTupleResult, 1, PyFloat_FromDouble( self->m_mae.eval() ) );
PyTuple_SET_ITEM( pyTupleResult, 2, PyFloat_FromDouble( self->m_rmse.eval() ) );
return pyTupleResult;
}