README.md

In marcusbuckmann/ffcr: Train two families of transparent classification models: fast-and-frugal trees and tallying

Fast-and-frugal classification in R (ffcr)

The ffcr package constructs two transparent classification models: fast-and-frugal trees and tallying models. The book Classification in the Wild: The Science and Art of Transparent Decision Making. (Katiskopoulos et al., 2020) describes these models, their applications and the algorithms to construct them in detail.

A fast-and-frugal tree is a decision tree with a simple structure: one branch of each node exits tree, the other continues to the next node until the final node is reached. A tallying model gives pieces of evidence the same weight. This package implements several methods for training these two models, ranging from simple heuristics to computationally more complex cross-entropy optimization (Rubinstein, 1999).

Installation

The package can be installed with the following command:

# install.packages("devtools")
devtools::install_github("marcusbuckmann/ffcr")

Training fast-and-frugal trees

We start by training a fast-and-frugal tree on a medical data set, predicting whether patients have liver disease (Ramana, Babu, and Venkateswarlu, 2011; Dua and Graff, 2017). The data set is included in the package.

library(ffcr)
model <- fftree(diagnosis~., data = liver, use_features_once = FALSE, method = "greedy", max_depth = 4)

The structure of the tree and its performance is shown by printing the model.

print(model)
#> Fast-and-frugal Tree object
#> Trained with : "recursive" method. 
#> 
#> Call: 
#> fftree(data = data, formula = formula, method = "greedy", max_depth = 4, 
#>     use_features_once = FALSE)
#> 
#> Formula: 
#> diagnosis ~ age + sex + totalBilirubin + directBilirubin + alkaline + 
#>     alamine + aspartate + proteins + albumin + albuminGlobulin
#> 
#> Tree: 
#> Reason / Prediction / (Proportion of class 'Liver disease') / (Number of objects classified)
#> 
#>   totalBilirubin > 1.65: Liver disease (0.91) (213)
#>     alkaline > 211.5: Liver disease (0.76) (132)
#>       age <= 25.5: No liver disease (0.30) (27)
#>         age > 25.5: Liver disease (0.55) (207)
#> 
#> 
#> Fitted values:
#>     Observed         Predicted        N  
#>     Liver disease    Liver disease    406
#>     No liver disease Liver disease    146
#>     Liver disease    No liver disease   8
#>     No liver disease No liver disease  19
#> 
#> Fitting performance:                            
#>     Accuracy            0.73
#>     Sensitivity         0.98
#>     Specificity         0.12
#>     Balanced accuracy   0.55
#>     F1 score            0.84
#>                   
#>     Depth     3.00
#>     Features  3.00
#>     Frugality 2.04

To visualize the tree we use

plot(model)

To make predictions according to the fast-and-frugal tree, we can use the predict function.

pred <- predict(model, newdata = liver[1:10,])
pred
#>  [1] "Liver disease"    "Liver disease"    "Liver disease"    "Liver disease"   
#>  [5] "Liver disease"    "Liver disease"    "Liver disease"    "Liver disease"   
#>  [9] "No liver disease" "Liver disease"

Training tallying models

To train tallying models, we use the tally function. To make predictions we use the same command as for the fast-and-frugal trees.

model <- tally(diagnosis~., data = liver, max_size = 4)
print(model)
#> Tallying object
#> Trained with : "basic" method. 
#> 
#> Call: 
#> tally(data = data, formula = formula, max_size = 4)
#> 
#> Formula: 
#> diagnosis ~ age + sex + totalBilirubin + directBilirubin + alkaline + 
#>     alamine + aspartate + proteins + albumin + albuminGlobulin
#> 
#> Reasons:                               
#>    + totalBilirubin   >    1.65
#>    + directBilirubin  >    0.85
#>    + aspartate        >   47.50
#>    + proteins         <=   3.65
#>    ____________________________________ 
#> Predict Liver disease if at least 1 reasons hold. 
#> 
#> 
#> Fitted values:
#>     Observed         Predicted        N  
#>     Liver disease    Liver disease    272
#>     No liver disease Liver disease     44
#>     Liver disease    No liver disease 142
#>     No liver disease No liver disease 121
#> 
#> Fitting performance:                            
#>     Accuracy            0.68
#>     Sensitivity         0.66
#>     Specificity         0.73
#>     Balanced accuracy   0.70
#>     F1 score            0.75
pred <- predict(model, newdata = liver[1:10,])

Please consult the vignette and the documentation of the package for more details on how to use it. The book Classification in the Wild provides background information on fast-and-frugal trees and tallying models.

References

Dua, Dheeru, and Casey Graff. 2017. “UCI Machine Learning Repository.” University of California, Irvine, School of Information; Computer Sciences. http://archive.ics.uci.edu/ml.

Katikopoulos, Konstantinos V., Özgür Şimşek, Marcus Buckmann, and Gerd Gigerenzer. 2020. Classification in the Wild: The Science and Art of Transparent Decision Making. MIT Press.

Ramana, Bendi Venkata, M Surendra Prasad Babu, and N. B. Venkateswarlu. 2011. “A Critical Study of Selected Classification Algorithms for Liver Disease Diagnosis.” International Journal of Database Management Systems 3 (2): 101–114.

Rubinstein, Reuven. 1999. “The Cross-Entropy Method for Combinatorial and Continuous Optimization.” Methodology and Computing in Applied Probability 1 (2): 127–190.

marcusbuckmann/ffcr documentation built on Jan. 4, 2024, 3:45 p.m.