Binary Tree Growth Algorithm for Feature Selection

Introduction

• This toolbox offers a Binary Tree Growth Algorithm ( BTGA )
• The Main file illustrates the example of how BTGA can solve the feature selection problem using benchmark data-set.

Input

• feat : feature vector ( Instances x Features )
• label : label vector ( Instances x 1 )
• N : number of trees
• max_Iter : maximum number of iterations
• N1 : number of trees in first group
• N2 : number of trees in second group
• N4 : number of trees in fourth group
• theta : tree reduction rate
• lambda : parameter controls nearest tree

Output

• sFeat : selected features
• Sf : selected feature index
• Nf : number of selected features
• curve : convergence curve

Example

% Benchmark data set 
load ionosphere.mat; 

% Set 20% data as validation set
ho = 0.2; 
% Hold-out method
HO = cvpartition(label,'HoldOut',ho);

% Parameter setting
N        = 10;
max_Iter = 100;
N1       = 3; 
N2       = 5;
N4       = 3;
theta    = 0.8;
lambda   = 0.5; 
% Binary Tree Growth Algorithm 
[sFeat,Sf,Nf,curve] = jBTGA(feat,label,N,max_Iter,N1,N2,N4,theta,lambda,HO);

% Plot convergence curve
plot(1:max_Iter,curve);
xlabel('Number of iterations');
ylabel('Fitness Value');
title('BTGA'); grid on;

Requirement

• MATLAB 2014 or above
• Statistics and Machine Learning Toolbox

Cite As

@article{too2018feature,
  title={Feature selection based on binary tree growth algorithm for the classification of Myoelectric Signals},
  author={Too, Jingwei and Abdullah, Abdul Rahim and Mohd Saad, Norhashimah and Mohd Ali, Nursabillilah},
  journal={Machines},
  volume={6},
  number={4},
  pages={65},
  year={2018},
  publisher={Multidisciplinary Digital Publishing Institute}
}