For feature selection of a binary class dataset, relief algorithm chooses an observation randomly and then selects nearest neighbors of that observation from both classes. Selecting number of nearest neighbors is an important parameter. I want to know how to choose that number.
you can use DistanceBand or KNN
you can check this
https://towardsdatascience.com/a-simple-introduction-to-k-nearest-neighbors-algorithm-b3519ed98e
One of the common practices is to use a supervised learning method where you train the system using a training set and then test classification results using an unknown test image. During the training phase using a trial and error approach you would need to observe which value of nearest neighbour gives the best results, and then use that value for the testing phase. However that is only valid for the dataset under study and may not be applicable globally.
- Badges
- Science topic