plot_ale

The plot_ale() function computes and visualizes
Accumulated Local Effects (ALE) for a selected feature, following the
approach described in Interpretable Machine Learning by Christoph Molnar.
ALE plots quantify how changes in a feature locally influence model
predictions, offering a robust alternative to Partial Dependence Plots (PDP)
by avoiding extrapolation and handling correlated predictors more reliably.

A minimal library specifically designed to make the estimation of Machine Learning
(ML) techniques as easy and accessible as possible, particularly within the framework of
the Knowledge Discovery in Databases (KDD) process in data mining. The package provides
essential tools to structure and execute each stage of a predictive or classification
modeling workflow, aligning closely with the fundamental steps of the KDD methodology,
from data selection and preparation, through model building and tuning, to the
interpretation and evaluation of results using Sensitivity Analysis. The 'MLwrap' workflow
is organized into four core steps; preprocessing(), build_model(), fine_tuning(), and
sensitivity_analysis(). It also includes global and pairwise interaction analysis based on
Friedman’s H-statistic to support a more detailed interpretation of complex feature
relationships.These steps correspond, respectively, to data preparation and transformation,
model construction, hyperparameter optimization, and sensitivity analysis. The user can
access comprehensive model evaluation results including fit assessment metrics, plots,
predictions, and performance diagnostics for ML models implemented through 'Neural
Networks', 'Random Forest', 'XGBoost' (Extreme Gradient Boosting), and 'Support Vector
Machines' (SVM) algorithms. By streamlining these phases, 'MLwrap' aims to simplify the
implementation of ML techniques, allowing analysts and data scientists to focus on
extracting actionable insights and meaningful patterns from large datasets, in line with
the objectives of the KDD process.

Albert Sesé

MLwrap

Machine Learning Modelling for Everyone

Javier Martínez García

Juan José Montaño Moreno

plot_ale function

<dl><dt>analysis_object</dt>
<dd>A fitted <code>wrap_object</code> with model results or
previously computed ALE values.</dd>
<dt>feature</dt>
<dd>Character. Name of the continuous feature for which ALE
should be computed.</dd>
<dt>group</dt>
<dd>Optional character. A grouping variable to compute grouped ALE
curves.</dd>
<dt>grid.size</dt>
<dd>Integer. Number of intervals to partition the feature domain
(default = 20).</dd>
<dt>use_test</dt>
<dd>Logical. If TRUE, ALE is computed using the test set
(default = FALSE).</dd>
<dt>plot</dt>
<dd>Logical. If TRUE, displays the ALE plot and returns <code>wrap_object</code>;
if FALSE, returns the ggplot object without modifying the object.</dd></dl>

Arguments

Plot Accumulated Local Effects (ALE) — plot_ale

<dl>

<dt>analysis_object</dt>
<dd>A fitted <code>wrap_object</code> with model results or
previously computed ALE values.</dd>


<dt>feature</dt>
<dd>Character. Name of the continuous feature for which ALE
should be computed.</dd>


<dt>group</dt>
<dd>Optional character. A grouping variable to compute grouped ALE
curves.</dd>


<dt>grid.size</dt>
<dd>Integer. Number of intervals to partition the feature domain
(default = 20).</dd>


<dt>use_test</dt>
<dd>Logical. If TRUE, ALE is computed using the test set
(default = FALSE).</dd>


<dt>plot</dt>
<dd>Logical. If TRUE, displays the ALE plot and returns <code>wrap_object</code>;
if FALSE, returns the ggplot object without modifying the object.</dd>

</dl>

plot_ale: Plot Accumulated Local Effects (ALE)

Description

Usage

Value

Arguments

References

See Also

Examples