3
$\begingroup$

I have a dataset of dimension 3,000 x 24,000 (approximately) with 6 class label. But the data is very sparse. The number of non-zero values per sample ranges from 10-300 (approx) out of 24,000. The non-zero values in the dataset are real numbers. I need to perform feature selection/reduction before the classification. Which technique would be better for such dataset?

$\endgroup$
  • $\begingroup$ Have you considered L1 regularization? $\endgroup$ – spdrnl Nov 29 '17 at 15:45
  • $\begingroup$ I am trying to apply elasticnet for feature selection. $\endgroup$ – Md. Abid Hasan Nov 29 '17 at 19:22
  • $\begingroup$ Elasticnet is L1 and L2; for feature selection L1 only offers better functionality AFAIK. $\endgroup$ – spdrnl Nov 30 '17 at 15:34
1
$\begingroup$

Feature selection and feature reduction are two very different strategies.

Generally speaking, non-parametric tests is probably your best option, I'd go towards a Kruskal-Wallis rank sum test to get overall differential features or a Mann-Whitney rank sum test for each class label.

For feature reduction, I believe zero-inflated factorial analysis (ZIFA) is a good solution (Pierson, Emma, and Christopher Yau. "ZIFA: Dimensionality reduction for zero-inflated single-cell gene expression analysis." Genome biology 16.1 (2015): 241.). However, more classical factor analysis, NMF or t-SNE may work.

Now the data you describe really looks a lot like a single-cell RNAseq dataset. If this is the case, I encourage you to take a look at the following resource: https://hemberg-lab.github.io/scRNA.seq.course/biological-analysis.html#de-in-a-real-dataset

$\endgroup$
  • $\begingroup$ I believe it is more correct to say factor analysis than factorial analysis, despite I don't know because didn't read those articles you cite. We say, factorial design, but that is very different. $\endgroup$ – ttnphns Nov 29 '17 at 11:44
  • $\begingroup$ Edit. Thanks for the comment, it was an incorrect translation. $\endgroup$ – Rémy Nicolle Nov 29 '17 at 13:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.