ref_probe_selection_twoStage: Two-stage feature selection

In jysonganan/methylDeConv: Computational pipeline for deconvolution of methylation data.

Two-stage feature selection

Description

Select features in two stages: firstly, select top features from one-vs-all t test; secondly, select the features with machine learning modeling.

Usage

ref_probe_selection_twoStage(
  ref_betamatrix,
  ref_phenotype,
  preselect = 300,
  ml_model = "elastic net"
)

Arguments

ref_betamatrix	The reference matrix ref_betamatrix.
ref_phenotype	The cell type information for the reference matrix.
preselect	The number of top features per cell type selected from one-vs-all t tests. The default value is 300.
ml_model	The machine learning model for feature selection in the second stage. The default value is "elastic net", which correpsonds to selecting the non-zero features from multi-class elastic net modeling on the reference matrix. Otherwise, if the parameter value is "RF", the model selection is based on the important variables learnt from multi-class Random forest modeling; if the parameter value is "rfe", it selects features based on recursive feature elimination algorithm and a Random Forest algorithm is used on each iteration to evaluate the model.

Value

Model class.

jysonganan/methylDeConv documentation built on Aug. 8, 2022, 6:25 a.m.