LASSO ( Least Absolute Shrinkage and Selection Operator) is a regularization method used in logistic regression to select the related features of a high-dimensional gene expression data sets. During the implementation using Python or R, I noticed that the selected features are different from those selected features in another round of implementation on the same data set.
First: Is this normal?
Second: how to choose the correct features if we repeat the above process several times?
Thanks in advance for your answers.
First, generally, if you need to get the same result, you need to fix the seed number value.
Second, the best way to evaluate the feature selection subset is to evaluate the selection process under, e.g., 10-fold cross-validation.
HTH.
Dr. Samer Sarsam
Samer Sarsam
Yes, I know that, but still you get different results.
Thanks for your answer.
Mohamed I. Riffi
In general, selecting the feature subset depends on the sample your working one and the search method your applying. For instance, in K-fold cross-validation, different subsets might be resulted from each fold. However, to get the final result of cross-validation (average result), you need to fix the seed number value.
Dr. Samer Sarsam
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python