Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 1

Shima Kamyab and Mahdi Eftekhari[1] studied various MO methods in feature selection problems.

They
proposed a binary version of existing evolutionary algorithms based on MO methods like Dynamic
Fitness Sharing(DFS), local best PSO variants, and a niching technique GA_SN_CM for selecting suitable
features from the available datasets. They compared the results with some well-known heuristic
approaches. In this work, the explanation of the proposed algorithm is missing.

They used the fitness function as

fitnessi ( selected features ) =¿

N t −N s
α ( classification accuracy ( selected features ) ) +(1−α )( ) (1)
Nt

Where Nt and Ns denote the number of total features and the number of selected features of a
candidate solution, respectively. α stand for a constant coefficient that makes a balance between
classification accuracy and number of selected features. The authors have used α as 0.7 in the
experiments. The authors used classification accuracy and number of selected features as trhe
evaluation criteria by each method. CatfishBPSO and Memetic Algorithm were used as comparison
algorithm. The authors compared Binary Genetic algorithm and Binary PSO with MO version of these
algorithms.

You might also like