Clustering Predictor Object

Description

A ClustPredictor object holds any unsupervised clustering algorithm and the data to be used for analyzing the model. The interpretation methods in the FACT package need the clustering algorithm to be wrapped in a ClustPredictor object.

Details

A Cluster Predictor object is a container for the unsupervised prediction model and the data. This ensures that the clustering algorithm can be analyzed in a robust way. The Model inherits from iml::Predictor Object and adjusts this Object to contain unsupervised Methods.

Super class

iml::Predictor -> ClustPredictor

Public fields

Methods

Public methods

• ClustPredictor$new()

• ClustPredictor$clone()

Method new()

Create a ClustPredictor object

Usage

ClustPredictor$new(
  model = NULL,
  data = NULL,
  predict.function = NULL,
  y = NULL,
  batch.size = 1000,
  type = NULL
)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

ClustPredictor$clone(deep = FALSE)

Arguments

Examples

require(factoextra)
require(FuzzyDBScan)
multishapes <- as.data.frame(multishapes[, 1:2])
eps = c(0, 0.2)
pts = c(3, 15)
res <- FuzzyDBScan$new(multishapes, eps, pts)
res$plot("x", "y")
# create hard label predictor
predict_part = function(model, newdata) model$predict(new_data = newdata, cmatrix = FALSE)$cluster
ClustPredictor$new(res, as.data.frame(multishapes), y = res$clusters,
                       predict.function = predict_part, type = "partition")
# create soft label predictor
predict_prob = function(model, newdata) model$predict(new_data = newdata)
ClustPredictor$new(res, as.data.frame(multishapes), y = res$results,
                               predict.function = predict_prob, type = "prob")

Idea - Isolated Effect on Assignment

Description

IDEA with a soft label predictor (sIDEA)
tacks changes the soft label of being assigned to each existing cluster throughout a (multidimensional) feature space IDEA with a hard label predictor (hIDEA)
tacks changes the soft label of being assigned to each existing cluster throughout a (multidimensional) feature space

Details

IDEA for soft labeling algorithms (sIDEA) indicates the soft label that an observation \textbf{x} with replaced values \tilde{\textbf{x}}_S is assigned to the k-th cluster. IDEA for hard labeling algorithms (hIDEA) indicates the cluster assignment of an observation \textbf{x} with replaced values \tilde{\textbf{x}}_S.

The global IDEA is denoted by the corresponding data set X:

\text{sIDEA}_X(\tilde{\textbf{x}}_S) = \left(\frac{1}{n} \sum_{i = 1}^n \text{sIDEA}^{(1)}_{\textbf{x}^{(i)}}(\tilde{\textbf{x}}_S), \dots, \frac{1}{n} \sum_{i = 1}^n \text{sIDEA}^{(k)}_{\textbf{x}^{(i)}}(\tilde{\textbf{x}}_S) \right)

where the c-th vector element is the average c-th vector element of local sIDEA functions. The global hIDEA corresponds to:

\text{hIDEA}_X(\tilde{\textbf{x}}_S) = \left(\frac{1}{n}\sum_{i = 1}^n \mathbb{1}_{1}(\text{hIDEA}_{\textbf{x}^{(i)}}(\tilde{\textbf{x}}_S)), \dots, \frac{1}{n}\sum_{i = 1}^n \mathbb{1}_{k}(\text{hIDEA}_{\textbf{x}^{(i)}}(\tilde{\textbf{x}}_S))\right)

where the c-th vector element is the fraction of hard label reassignments to the c-th cluster.

Public fields

Active bindings

Methods

Public methods

• IDEA$new()

• IDEA$plot()

• IDEA$plot_globals()

• IDEA$clone()

Method new()

Create an IDEA object.

Usage

IDEA$new(predictor, feature, method = "g+l", grid.size = 20L, noise.out = NULL)

Arguments

Returns

(data.frame)
Values for the effect curves:
One row per grid per instance for each local idea estimation. If method includes global estimation, one additional row per grid point.

Method plot()

Plot an IDEA object.

Usage

IDEA$plot(c = NULL)

Arguments

Returns

(ggplot)
A ggplot object that depends on the method chosen.

Method plot_globals()

Plot the global sIDEA curves of all clusters.

Usage

IDEA$plot_globals(mass = NULL)

Arguments

Returns

(ggplot)
A ggplot object.

Method clone()

The objects of this class are cloneable with this method.

Usage

IDEA$clone(deep = FALSE)

Arguments

See Also

iml::FeatureEffects, iml::FeatureEffects

Examples

# load data and packages
require(factoextra)
require(FuzzyDBScan)
multishapes = as.data.frame(multishapes[, 1:2])
# Set up an train FuzzyDBScan
eps = c(0, 0.2)
pts = c(3, 15)
res = FuzzyDBScan$new(multishapes, eps, pts)
res$plot("x", "y")
# create soft label predictor
predict_prob = function(model, newdata) model$predict(new_data = newdata)
predictor = ClustPredictor$new(res, as.data.frame(multishapes), y = res$results,
                                    predict.function = predict_prob, type = "prob")
# Calculate `IDEA` global and local for feature "x"
idea_x = IDEA$new(predictor = predictor, feature = "x", grid.size = 5)
idea_x$plot_globals(0.5) # plot global effect of all clusters with 50 percent of local mass.

SMART - Scoring Metric after Permutation

Description

SMART estimates the importance of a feature to the clustering algorithm by measuring changes in cluster assignments by scoring functions after permuting selected feature. Cluster-specific SMART indicates the importance of specific clusters versus the remaining ones, measured by a binary scoring metric. Global SMART assigns importance scores across all clusters, measured by a multi-class scoring metric. Currently, SMART can only be used for hard label predictors.

Details

Let M \in \mathbb{N}_0^{k \times k} denote the multi-cluster confusion matrix and M_c \in \mathbb{N}_0^{2 \times 2} the binary confusion matrix for cluster c versus the remaining clusters. SMART for feature set S corresponds to:

\text{Multi-cluster scoring:} \quad \text{SMART}(X, \tilde{X}_S) = h_{\text{multi}}(M) \\ \text{Binary scoring:} \quad \text{SMART}(X, \tilde{X}_S) = \text{AVE}(h_{\text{binary}}(M_1), \dots, h_{\text{binary}}(M_k))

where \text{AVE} averages a vector of binary scores, e.g., via micro or macro averaging. In order to reduce variance in the estimate from shuffling the data, one can shuffle t times and evaluate the distribution of scores. Let \tilde{X}_S^{(t)} denote the t-th shuffling iteration for feature set S. The SMART point estimate is given by:

\overline{\text{SMART}}(X, \tilde{X}_S) = \psi\left(\text{SMART}(X, \tilde{X}_S^{(1)}), \dots, \text{SMART}(X, \tilde{X}_S^{(t)})\right)

where \psi extracts a sample statistic such as the mean or median or quantile.

Public fields

Methods

Public methods

• SMART$new()

• SMART$print()

• SMART$plot()

• SMART$clone()

Method new()

Create a SMART object

Usage

SMART$new(
  predictor,
  features = NULL,
  metric = "f1",
  avg = NULL,
  n.repetitions = 5
)

Arguments

Returns

(data.frame)
data.frame with the results of the feature importance computation. One row per feature with the following columns: For global scores:

• importance.05 (5% quantile of importance values from the repetitions)

• importance (median importance)

• importance.95 (95% quantile) and the permutation.error (median error over all repetitions). For cluster specific scores each column indicates for a different cluster.

Method print()

Print a SMART object

Usage

SMART$print()

Returns

character
Information about predictor, data, metric, and avg and head of the results.

Method plot()

plots the similarity score results of a SMART object.

Usage

SMART$plot(log = FALSE, single_cl = NULL)

Arguments

Details

The plot shows the similarity per feature. For global scores: When n.repetitions in SMART$new was larger than 1, then we get multiple similarity estimates per feature. The similarity are aggregated and the plot shows the median similarity per feature (as dots) and also the 90%-quantile, which helps to understand how much variance the computation has per feature. For cluster-specific scores: Stacks the similarity estimates of all clusters per feature. Can be used to achieve a global estimate as a sum of cluster-wise similarities.

Returns

ggplot2 plot object

Method clone()

The objects of this class are cloneable with this method.

Usage

SMART$clone(deep = FALSE)

Arguments

See Also

iml::FeatureImp

SMART

SMART

Examples

# load data and packages
require(factoextra)
require(FuzzyDBScan)
multishapes = as.data.frame(multishapes[, 1:2])
# Set up an train FuzzyDBScan
eps = c(0, 0.2)
pts = c(3, 15)
res = FuzzyDBScan$new(multishapes, eps, pts)
res$plot("x", "y")
# create hard label predictor
predict_part = function(model, newdata) model$predict(new_data = newdata, cmatrix = FALSE)$cluster
predictor = ClustPredictor$new(res, as.data.frame(multishapes), y = res$clusters,
                               predict.function = predict_part, type = "partition")
# Run SMART globally
macro_f1 = SMART$new(predictor, n.repetitions = 50, metric = "f1", avg = "macro")
macro_f1 # print global SMART
macro_f1$plot(log = TRUE) # plot global SMART
# Run cluster specific SMART
classwise_f1 = SMART$new(predictor, n.repetitions = 50, metric = "f1")
macro_f1 # print regional SMART
macro_f1$plot(log = TRUE) # plot regional SMART

Create a generic prediction function

Description

Create the algorithms prediction function.

Usage

create_predict_fun(model, task, predict.fun = NULL, type = NULL)

## S3 method for class 'Learner'
create_predict_fun(model, task, predict.fun = NULL, type = NULL)

Arguments

function(model, newdata)

Value

A unified cluster assignment function for either hard or soft labels.

Methods (by class)

• create_predict_fun(Learner): Create a predict function for algorithms from mlr3cluster

Evaluate Class

Description

Calculation of binary similarity metric based on confusion matrix.

Usage

evaluate_class(actual, predicted, metric = "f1")

calculate_confusion(actual, predicted)

Arguments

Value

A binary score for each of the clusters and the number of instances.

Functions

• calculate_confusion(): Calculate confusion matrix