demo10.cpp File Reference

Example 10: Basic Filter-based feature selection. More...

#include <memory>
#include <exception>
#include <iostream>
#include <cstdlib>
#include <string>
#include <vector>
#include "error.hpp"
#include "global.hpp"
#include "subset.hpp"
#include "data_intervaller.hpp"
#include "data_splitter.hpp"
#include "data_splitter_5050.hpp"
#include "data_scaler.hpp"
#include "data_scaler_void.hpp"
#include "data_accessor_splitting_memTRN.hpp"
#include "data_accessor_splitting_memARFF.hpp"
#include "criterion_normal_gmahalanobis.hpp"
#include "distance_euclid.hpp"
#include "classifier_knn.hpp"
#include "seq_step_straight.hpp"
#include "search_seq_sfs.hpp"

Functions
int	main ()

Detailed Description

Example 10: Basic Filter-based feature selection.

Function Documentation

main()

int main

(

)

Example 10: Basic Filter-based feature selection.

This is a simple example of filter-based feature selection, as described in the seminal book by Devijver and Kittler from 1982 - Pattern Recognition: A Statistical Approach. Features are selected here using the Sequential Forward Selection (SFS) procedure so as to maximize the Generalized Mahalanobis probabilistic class distance based on the assumption of normality of the data. Generalized Mahalanobis is evaluated on the first 50% of data samples. The selected subset is eventually verified by means of 3-NN classifier accuracy estimation on the second 50% (independent test) part of the data. SFS is called here in d-parametrized setting, invoked by nonzero parameter d in search(d,...). In this scenario the user has to decide about the target subset size.

Note

With arbitrary data the assumption of normality may not be fulfilled what would negatively affect the feature selection results based on Mahalanobis, Bhattacharyya or Divergence.

Examples

demo10.cpp, demo52t.cpp, and demo55.cpp.