CODEWITHSUNDEEP
machine-learningclassificationfeature-extractionfeature-selection
BioChemoinformatics
BioChemoinformatics

Reputation: 715

How to combine two (or multiple) kinds of features as one final feature to build classification model?

Currently, I meeting such question:How to combine two (or multiple) kinds of features as one final feature to build classification model?

For example, I would like to do a classification model to predict the drug-target interaction, here for each drug I can get 500 features, and each targets I can get 800 features (this two kinds of features are independent each other). As a know, a simple method to combine this two kinds of features which just plus two together (that is 500 + 800 = 1300 feature for each drug-target pair).

Does anybody know other methods to do such kind of thing and use the combined feature to build classification models?

Upvotes: 2

Views: 5331

Answers (2)

dkohlsdorf
dkohlsdorf

Reputation: 1

Random Forest select the best features for your classification task using information gain. The classifier works for multiple feature sources and also types. For example, you can combine continuous attributes and discrete ones.

The training time is a little longer since you have to iterate all features multiple times but the memory performance and the classification speed is pretty good.

Upvotes: 0

Has QUIT--Anony-Mousse
Has QUIT--Anony-Mousse

Reputation: 77454

There are a number of methods which should work just fine even with this many features, such as SVM.

Also, there is a whole bunch of literature on this, including SVD, PCA, MDS, feature selection, feature transformation... you'll have to read up on these, we can't pick your magic bullet out of all that for you without having your data.

Upvotes: 3

Related Questions