Cosan::MissingValues< NumericType > Class Template Reference

#include <missingvalues.h>

Inheritance diagram for Cosan::MissingValues< NumericType >:

Public Member Functions
	MissingValues ()=delete
	MissingValues (CosanRawData< NumericType > &RD)
Public Member Functions inherited from Cosan::Preprocessor< NumericType >
	Preprocessor ()
virtual	~Preprocessor ()=default
virtual void	fit (const CosanMatrix< NumericType > &X)
virtual CosanMatrix< NumericType >	transform (const CosanMatrix< NumericType > &X)
Public Member Functions inherited from Cosan::CosanBO
	CosanBO ()
	Default constructor. More...
virtual const std::string	GetName () const
	Get the name of the objects. More...

Private Member Functions
void	fit (CosanRawData< NumericType > &RD)

Detailed Description

template<Numeric NumericType>
class Cosan::MissingValues< NumericType >

Definition at line 13 of file missingvalues.h.

Constructor & Destructor Documentation

MissingValues() [1/2]

template<Numeric NumericType>

Cosan::MissingValues< NumericType >::MissingValues ( )

delete

MissingValues() [2/2]

template<Numeric NumericType>

Cosan::MissingValues< NumericType >::MissingValues ( CosanRawData< NumericType > &  RD )

inline

Definition at line 16 of file missingvalues.h.

                                                        :Preprocessor<NumericType>(){
                this->fit(RD);
            }

Member Function Documentation

fit()

template<Numeric NumericType>

void Cosan::MissingValues< NumericType >::fit ( CosanRawData< NumericType > &  RD )

inlineprivate

Definition at line 21 of file missingvalues.h.

                                                    {
                fmt::print("*********************************\n");
                fmt::print("Begin cleaning up missing data!\n");
                std::set<gsl::index> MissingIdxX;
                for (auto & each: RD.GetIdxmissingX()){
                    MissingIdxX.insert(RD.GetRawToNumIdx()[each[1]]);
                }
                bool HasMissingIdxY = RD.GetIdxmissingY().size()==0 ? false :true;
                CosanMatrix<NumericType> X = RD.GetInput();
 
                for (auto &i : MissingIdxX ){
                    gsl::index _numofMissing = X.col(i).array().isNaN().template cast<NumericType>().sum();
                    fmt::print("For X Column {:d} has {:d} missing value(s).\n",i,_numofMissing);
                    NumericType mean = X.col(i).unaryExpr([](NumericType v) { return std::isnan(v)? 0 : v; }).sum()/(X.rows()-_numofMissing);
                    fmt::print("Fill the missing/nan entry with the column mean={:f}.\n",mean);
                    X.col(i) = X.col(i).array().isNaN().select(0,X.col(i)).array()+mean*(X.col(i).array().isNaN().template cast<NumericType>());
 
                }
                if (HasMissingIdxY==true){
                    gsl::index i = 0;
                    CosanMatrix<NumericType> Y = RD.GetTarget();
                    gsl::index _numofMissing = Y.col(i).array().isNaN().template cast<NumericType>().sum();
                    fmt::print("For Y has {:d} missing value(s).\n",_numofMissing);
 
                    NumericType mean = Y.col(i).unaryExpr([](NumericType v) { return std::isnan(v)? 0 : v; }).sum()/(Y.rows()-_numofMissing);
                    fmt::print("Fill the missing/nan entry with the column mean={:f}.\n",mean);
                    Y.col(i) = Y.col(i).array().isNaN().select(0,Y.col(i)).array()+mean*(Y.col(i).array().isNaN().template cast<NumericType>());
                    RD.UpdateData(X,Y);
                }
                else{
                    RD.UpdateData(X);
                }
                fmt::print("End of cleaning process\n");
                fmt::print("*********************************\n");
            }

---

The documentation for this class was generated from the following file:

• cosan/preprocessing/missingvalues.h