Hi James,

PSO algorithm works by having a population (ie., swarm) of candidate solutions (particles). These particles are moved around in the search-space. The movements of the particles are controlled by their best known position in the space and the entire swarm's best known position. When improved positions are being discovered these will then come to guide the movements of the swarm. The process is repeated several times in order to (hopefully) obtain a satisfactory solution. After performing the search task, one, for instance, can evaluate the worth of a subset of attributes by considering the individual predictive ability of each feature along with the degree of redundancy between them. This work needs to be done under the evaluation of cross-validation, for example.

HTH.

Samer, PhD.

Hi James,

Feature selection is an optimization problem where the objective is to select the minimum number of features that have the maximum informativeness. To apply PSO to the feature selection problem you need first to map features selection/deselection using a representation suitable for PSO (usually continuous values representing particle's position), develop the particles evaluation function, generate the initial swarm, and repeatedly apply the PSO steps of particles evaluation and update their velocity and positions till a predefined stopping criteria is met.

Anwar

I can use the comment of Anwar and I can add:

- which kind of problem are you trying to solve?

- you need to define some fitness function to compare the particles

Dear Josimar Chire

I have 20 features extracted from a signal and i want to select the features that able to improve the classification accuracy. May i know is that by applying the "mean" algorithms in PSO can solve above situation?

Thank you

First, I suggest to analyze variance of your dimensions then you clean your data, less dimensions.

After, you are looking for the best combination to get the maximum accuracy. Then, you need to model your problem following this criterion.

Check this:

https://www.neuraldesigner.com/blog/genetic_algorithms_for_feature_selection

You can write me a private message.

Visit link for Matlab Code

https://youtu.be/OQ3T575sbI8

http://dg-algorithm.blogspot.com/

I use the dataset from UCI repository. Say for diabetes dataset, the features forms the dimensions in PSO. If so, If I initialize the particles with these feature values of the dataset, then if the particles move, then I'm losing the patient's data. How to perform feature selection using PSO is my question. Can someone suggest for medical data.

I don't have an idea about how to code pso for feature selection, can anybody help me please.

For this, find the answer in the following paper

Article Selection and classification of gene expression in autism di...

Dear Shilan

i wants to know how we get the accuracy after applying the PSO i read you article. i have theoretical concept but i am stuck in matlab code

Just take a look at the video explanation:

https://youtu.be/EHnTdY40_6A

You may check out this python code

https://github.com/CMATER-JUCS/Py_FS/blob/main/Py_FS/wrapper/nature_inspired/PSO.py

You can have a GitHub repo from the Jadavpur University CS department:

https://github.com/CMATER-JUCS/Py_FS/blob/main/Py_FS/wrapper/nature_inspired/PSO.py

BTW there are many other useful OA codes available at https://github.com/CMATER-JUCS/Py_FS