Filters for survival analysis

[mbostanara]

I want to run a number of machine learning algorithms with different feature selection methods on survival data using the MLR3 package. For that, I am using the Benchmark() function of MLR3.

Unfortunately, filter feature selection methods of MLR3 do not support survival, yet. However, MLR package supports survival filters.

I can fuse MLR learners with an MLR filter method. After that, I need to convert them to a learner in MLR3 in order to be able to use banchmark_grid() function of MLR3.

I would like to know if there is any way to use MLR survival filters in MLR3? Or is there any way that I can convert MLR filters to MLR3 filters?

Thank you

[RaphaelS1]

Hi, I'll let someone else answer the direct question but I think the deeper point is that we need to implement the surv task type in mlr3filters. I can try look into this when I have some time but that's quite limited now, can I suggest you open an issue in mlr3filters requesting support for surv task type?

[mllg]

A few comments:

1. You can already apply some filters on survival data, i.e. importance filters (flt("importance")) and univariate model scores (flt("performance")). Importance currently is only supported by random forests and xgboost, we are looking into implementing it into more models (such as CoxPH).
2. There is this PR to at least support a simple variance filter: Allow NA as task type mlr3filters#106
3. AFAICT mlr supports only one survival filter which we currently do not support: mRMRe.
4. https://doi.org/10.1093/bib/bbab354 uses martingale residuals to transform the survival task to a regression task and eventually be able to apply a broad range of regression filters on it, this is definitely something we can think about to also support in the future.

You can find a some examples on how to use filters in a pipeline in our gallery, e.g. here: https://mlr3gallery.mlr-org.com/posts/2020-07-27-bikesharing-demand/

Does this help?

[mbostanara]

Hi @RaphaelS1, Thank you for the response. That is right. I will surely request that.

[mbostanara]

Hi @mllg,
Thank you for the comments. I will check them out.