Model performance audit"

knitr::opts_chunk$set(warning = FALSE)
knitr::opts_chunk$set(message = FALSE)

Data

To illustrate application of auditor we will use dataset "dragons" available in the DALEX package. The dataset contains characteristics of fictional creatures (dragons), like year of birth, height, weight, etc (see below). The goal is to predict the length of life of dragons (a regression problem).

library(DALEX)
data(dragons)
head(dragons)

Models

First, we need models to compare. We selected linear regression and random forest because of their different structures. Linear regression model linear relationships between target response and independent variables. While random forest should be able to capture also non-linear relationships between variables.

# Linear regression
lm_model <- lm(life_length ~ ., data = dragons)

# Random forest
library(randomForest)
set.seed(59)
rf_model <- randomForest(life_length ~ ., data = dragons)

Preparation for analysis of performance

Analysis begins with creation of an explainer object with explain function from DALEX package. Explainer wraps a model with its meta-data, such as dataset that was used for training or observed response.

lm_exp <- DALEX::explain(lm_model, label = "lm", data = dragons, y = dragons$life_length)
rf_exp <- DALEX::explain(rf_model, label = "rf", data = dragons, y = dragons$life_length)

Next step requires creation of model_performance objects of each explained model. The function computes chosen performance measures for passed model.

library(auditor)
lm_mp <- model_performance(lm_exp)
rf_mp <- model_performance(rf_exp)

lm_mp

Model ranking radar plot

Model performance measures may be plotted together to easily compare model performances. A result further from the center of the plot means a better model performance. Parameter verbose indicates whether the table with scores should be generated. On the plot scores are inversed and scaled to [0,1].

plot(lm_mp, rf_mp)
# alternative:
# plot_radar(lm_mp, rf_mp, verbose = FALSE)

There is a possibility to define functions with custom model performance measure.

new_score <- function(object) sum(sqrt(abs(object$residuals)))

lm_mp <- model_performance(lm_exp,  
                          score = c("mae", "mse", "rec", "rroc"), 
                          new_score = new_score)

rf_mp <- model_performance(rf_exp,  
                          score = c("mae", "mse", "rec", "rroc"), 
                          new_score = new_score)

plot(lm_mp, rf_mp)

Other methods

Other methods and plots are described in vignettes:



Try the auditor package in your browser

Any scripts or data that you put into this service are public.

auditor documentation built on July 27, 2021, 1:08 a.m.