Know–Guided Regularized Random Forests
A general framework to use prior information to guide feature selection in the framework of random forests

  • Use prior information from a single domain or from multiple domains.
  • Combine prior information from multiple domains in a linear or non-linear form.
  • Optimize the weight given to each type of domain knowledge concurrently with feature selection procedure.
  • The method is available as the KnowGRFF package in the R/CRAN archive (https://cran.r-project.org/web/packages/KnowGRRF/index.html)
Citation:

  • Guan X, Liu L. (2018) Know-GRRF: Domain-Knowledge Informed Biomarker Discovery with Random Forests. 6th International Work-conference on Bioinformatics and Biomedical Engineering. Granada, Spain.
  • Guan X, Runger G, Liu L (2019) Dynamic Incorporation of Prior Knowledge from Multiple Domains in Biomarker Discovery. BMC Bioinformatics (in press).