RSRF

Description

Random Splitting Random Forest algorithms for mixed data including functional and scalar data. Response Types

• Continuous
• Categorical
• Survival

Version 1.0

Origin

It is an extension of the Möller et al. (2016) algorithm with the following new options:

• Multiple covariates
• Multiple functioanl covariates and non-functioanl covariates
• The respose type are Survival with censoring, Categorical and Continuous.
• The Variable Importance Plot (VIP) is enhanced.
• The random forest engine is from randomForestSRC by Ishwaran et al (2022).

Install in R

You can install the RSRS package from github and use it directly in R or R-Studio:

library(devtools)
install_github("mohammad-fayaz/RSRF")

or

library(pak)
pak::pkg_install("mohammad-fayaz/RSRF")

Vignettes

It has one vignette in this version:

• Random Splitting Random Forest for Categorical Response You can run it step by step.

It is in the Vignettes folder. (PDF file 13 pages) Vignettes - Link.

References

• [Oral Presentation] Fayaz M. and Abadi A.,Functional Random Forest for Mixed Data, The 41st Annual Conference of the International Society for Clinical Biostatistics (ISCB 2020)
• [Poster Presentation] Fayaz M., Shakeri N., Abadi A. and Khodakarim S., Random splitting random forest for survival analysis with non-functional and functional covariates in the EEG-fNIRS trial, The 14th International Conference of the ERCIM WG on Computational and Methodological Statistics (CMStatistics 2021)
• [Under Review] A Paper is submitted to the Peer-reviewed journal.
• Möller, Annette, Gerhard Tutz, and Jan Gertheiss. "Random forests for functional covariates." Journal of Chemometrics 30, no. 12 (2016): 715-725.
• Ishwaran, Hemant, Udaya B. Kogalur, and Maintainer Udaya B. Kogalur. "Package ‘randomForestSRC’." breast 6 (2022): 1.

Contacts

Mohammad Fayaz, PhD in Biostatistics ORCID: 0000-0002-5643-9763