is it possible to do feature selection for unsupervised machine learning problems?

I started looking for ways to do feature selection in machine learning.

By having a quick look at this post , I made the assumption that feature selection is only manageable for supervised learning problems:

Still, I have to ask: are there methods to do feature selection without having a known variable that will be used for a classification/regression problem?

• You're missing the link to the post you're talking about. Better than putting the link, you should put an extract of the part that is meaningful to your question. – MaxouMask Mar 27 '18 at 9:30

Feature Selection is a technique which is used when we you know the target variable(Supervised Learning)

When we talk with respect to Unsupervised Learning, there is no exact technique which could do that. But there is something which can help us in those lines i.e., Dimensionality Reduction, this technique is used to reduce the number of features and give us the features which explains the most about the dataset. The features would be derived from the existing features and might or might not be the same features.

There are different techniques which are available for doing so:

1. PCA
2. Linear discriminant analysis
3. Non-negative Matrix Factorization
4. Generalized discriminant analysis and many more.

The outcome of Feature Selection would be the same features which explain the most with respect to the target variable but the outcome of the Dimensionality Reduction might or might not be the same features as these are derived from the given input.

There are some methods to feature selection on unsupervised scenario:

• Laplace Score feature selection;
• Spectral Feature selection
• GLSPFS feature selection;
• JELSR feature selection
• Filter low-variance features (an implementation here)
• Filter correlated features (can be implemented using corr() from pandas)