README.md

In PierreMasselot/primr: Patient rule-induction method

primr

Patient rule-induction method

Description

The primr package provides function for performing the patient rule-induction method (PRIM) proposed by Friedman and Fisher (1999). PRIM is designed for bump hunting, i.e. to find a subdomain of x inputs in which an objective function of a response y is high.

Installation

1. In R, install the package directly from github using the command (the package devtools is required):

> library(devtools)
> install_github("PierreMasselot/primr", build_vignettes = TRUE)

1. The package can then be loaded as usual: library(primr).
2. You can see the vignette for simple examples: vignette("toy_example").
3. You can see the list of functions below and access help from R with ?peeling.

Functions

The primr package revolves around two main functions : peeling: Performs the top-down peeling consisting by iteratively peeling a box containing the whole dataset such that the objective function increases. pasting: Refines the final box's edges by slightly expanding it, increasing the objective function value.

Both function produce a prim object that contains the peeling trajectory, i.e. the successive peeled boxes. The stopping box of the peeling algorithm can be chosen through different functions: jump.prim: Selects the stopping box in a prim object through a 'jump' criterion. cv.trajectory: Produces a cross-validated peeling trajectory. * plot_trajectory: Plots the peeling trajectory.

In addition, prim objects can be passed to several functions for analysis: extract.box: Extracts a particular box from a prim object. plot_box: Plots a bidimensional projection of the data with one or several boxes. * predict.prim: For a new set of data, predicts whether each observation falls in the chosen box.

References

Friedman, J.H., Fisher, N.I., 1999. Bump hunting in high-dimensional data. Statistics and Computing 9, 123-143. https://doi.org/10.1023/A:1008894516817

Masselot P., Chebana F., Campagna C., Lavigne E., Ouarda T.B.M.J., Gosselin P. Machine learning approaches to identify thresholds in a heat-health warning system context. Submitted.

PierreMasselot/primr documentation built on Feb. 5, 2021, 7:33 p.m.