Paper Author : Yan Shuo Tan, Chandan Singh, Keyan Nasseri, Abhineet Agarwal, Bin Yu
R Package Author: Haoxue Wang ([email protected])---University of Cambridge
This package is R version for the FIGS algorithms based on python, hopefully following R version for imodels will be developed in the furture.
The introduction manual of the package is in Manual. The plot function is still TBD.
Fast Interpretable Greedy-Tree Sums (FIGS) is an algorithm for fitting concise rule-based models. Specifically, FIGS generalizes CART to simultaneously grow a flexible number of trees in a summation. The total number of splits across all the trees can be restricted by a prespecified threshold, keeping the model interpretable. Experiments across real-world datasets show that FIGS achieves state-of-the-art prediction performance when restricted to just a few splits (e.g. less than 20). 'FIGS' is first defined in Tan et al. (2022) https://arxiv.org/abs/2201.11931
figs.regressor: fitting function for fast interpretable greedy-tree sums on regression
figs.classifier: fitting function for fast interpretable greedy-tree sums on regression
install.packages("figs") # not available as still in revison and not published yet
install.packages("rpart") # based on orginal tree split algorithms
install.packages("fastshap") # to generate friedman dataset as required
library(rpart)
library(fastshap) # load friedman datasetsource("fit.R") # please change the working environment as belowed first
# getwd() # get the current directory
# setwd("xxxx") # set the path as working environment
# you can reproduce the simulation conveniently by loading the original data
# load("data_simu.RData") # under the document of figs/inst/extdata
# load the data for classification and regression
# X_reg = gen_friedman(100)[,-1]
# Y_reg = data.frame(gen_friedman(100)[,1])
# Y_reg = Y_reg$gen_friedman.100....1.
data_reg = read.csv("data_reg.csv") # under the document of figs/inst/extdata
data_cls <- read.csv(url("https://archive.ics.uci.edu/ml/machine-learning-databases/breast-cancer-wisconsin/wdbc.data"), header = FALSE)
X_cls = data_cls[,3:32] # load breast_cancer dataset for classification
Y_cls = data_cls[,2]
fit1 <- figs.classifier(X_cls, Y_cls, max_rules=6, sample_weight = rep(1,nrow(X_cls)))
fit2 <-figs.regressor(X_reg, Y_reg, max_rules=12, sample_weight = rep(1,nrow(X_cls)))
> fit1 <- figs.classifier(X_cls, Y_cls, max_rules=3, sample_weight = rep(1,nrow(X_cls)))
adding node V23 <= 16.795 Val: 0.9129288 (left) 0.05789474 (right) # the same result as the python simulation
adding node V30 <= 0.1358 Val: 0.3913043 (left) 0.984985 (right) # for the classification problem, we transfer y into number for residual caculation
adding node V25 <= 104.95 Val: 0.04166667 (left) 0.9090909 (right)
> fit2 <-figs.regressor(X_reg, Y_reg, max_rules=11, sample_weight = rep(1,nrow(X_cls)))
adding node x6 <= 0.4455641 Val: 13.38568 (left) 16.10976 (right) # see the output in R for detail structure
adding node x1 <= 0.4043588 Val: 11.91155 (left) 15.26184 (right)
adding node x2 <= 0.4892177 Val: 12.82209 (left) 18.30681 (right)
adding node x9 <= 0.2732258 Val: 11.72698 (left) 16.11151 (right)
adding node x4 <= 0.791582 Val: 11.96541 (left) 16.78442 (right)
adding node x2 <= 0.4435662 Val: 13.34833 (left) 17.1886 (right)
adding node x9 <= 0.654545 Val: 14.72727 (left) 19.95271 (right)
adding node x5 <= 0.790525 Val: 12.67281 (left) 17.07141 (right)
adding node x1 <= 0.4024863 Val: -1.136916 (left) 1.092331 (right) # begin to grow another tree to fit the data
adding node x2 <= 0.1795486 Val: 11.02559 (left) 16.95749 (right)
adding node x10 <= 0.3362207 Val: 12.2713 (left) 16.47628 (right)
> 
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent