cmaRs: A powerful predictive data mining package in R

thumbnail.default.placeholder
Date
2023-12-15
Authors
Yerlikaya-Özkurt, Fatma
Yazıcı, Ceyda
Batmaz, İnci
Journal Title
Journal ISSN
Volume Title
Publisher
SoftwareX
Abstract
Conic Multivariate Adaptive Regression Splines (CMARS) is a very successful method for modeling nonlinear structures in high-dimensional data. It is based on MARS algorithm and utilizes Tikhonov regularization and Conic Quadratic Optimization (CQO). In this paper, the open-source R package, cmaRs, built to construct CMARS models for prediction and binary classification is presented with illustrative applications. Also, the CMARS algorithm is provided in both pseudo and R code. Note here that cmaRs package provides a good example for a challenging implementation of CQO based on MOSEK solver in R environment by linking R to MOSEK through the package Rmosek.
Description
Open Access, Published by SoftwareX, https://doi.org/10.1016/j.softx.2023.101553, Fatma Yerlikaya-Özkurt, Department of Industrial Engineering, Atılım University, Ankara, Turkey, Ceyda Yazıcı, Department of Mathematics, TED University, Ankara, Turkey, İnci Batmaz, Department of Statistics, Middle East Technical University, Ankara, Turkey.
Keywords
Conic multivariate adaptive regression splines, Nonparametric regression, Tikhonov regularization, Conic quadratic programming, Interior point method, Binary classification
Citation
http://hdl.handle.net/20.500.14411/1927