Weighted Nearest Neighbour Imputation. More...

#include <yat/utility/WeNNI.h>

Inheritance diagram for theplu::yat::utility::WeNNI:

Public Member Functions
	WeNNI (const utility::Matrix &matrix, const utility::Matrix &weight, const unsigned int neighbours)
unsigned int	estimate (void)
	Function doing WeNNI imputation.
const utility::Matrix &	imputed_data_raw (void) const
const utility::Matrix &	imputed_data (void) const
const std::vector< size_t > &	not_imputed (void) const
Protected Member Functions
std::vector< std::pair< size_t, double > >	calculate_distances (const size_t) const
std::vector< size_t >	nearest_neighbours (const size_t, const std::vector< std::pair< size_t, double > > &) const
Protected Attributes
const utility::Matrix &	data_
utility::Matrix	imputed_data_
unsigned int	neighbours_
std::vector< size_t >	not_imputed_
const utility::Matrix &	weight_

Detailed Description

Weighted Nearest Neighbour Imputation.

WeNNI is a continuous weights generalization of the (binary weights) kNNI algorithm presented by Troyanskaya et al. A reference to this paper is found in the NNI document referred to in the NNI class documentation. The NNI document also describes WeNNI in depth.

Note:: Missing values should be represented with a zero weight. WeNNI will treat the corresponding data values as zero, i.e., this implies that NaNs and Infs with zero weight will not have any impact on calculations.a

See also:: NNI and kNNI

Constructor & Destructor Documentation

theplu::yat::utility::WeNNI::WeNNI	(	const utility::Matrix &	matrix,
		const utility::Matrix &	weight,
		const unsigned int	neighbours
	)

Constructor

Member Function Documentation

std::vector<std::pair<size_t,double> > theplu::yat::utility::NNI::calculate_distances ( const size_t ) const

protectedinherited

$d_{ij}^2=\frac {\sum_{k=1}^C w_{ik} w_{jk} (x_{ik}-x_{jk})^2 }{\sum_{k=l}^C w_{ik} w_{jk} }$ where C is the number of columns

unsigned int theplu::yat::utility::WeNNI::estimate ( void )

virtual

Function doing WeNNI imputation.

Perform WeNNI on data in matrix with continuous uncertainty weights in weight using neighbours for the new impute value.

The return value can be used as an indication of how well the imputation worked. The return value should be zero if proper pre-processing of data is done. An example of bad data is a matrix with a column of zero weights, another is a corresponding situation with a row with all weights zero.