SONO (Scores Of Nominal Outlyingness)

The SONO (Scores Of Nominal Outlyingness) R package includes a function that can be used for detecting outliers in data sets consisting of nominal data. Some of the capabilities of the package include:

• Calculating scores of outlyingness for data sets consisting of nominal variables.
• Estimating the maximum length of nominal sequences (MAXLEN) for doing frequent pattern mining.
• Computing maximum/minimum itemset support threshold values.
• Visualising the matrix of variable contributions to the score of nominal outlyingness computed.
• Computing evaluation metrics to compare performance of outlier detection algorithms (ROC AUC, Recall@K, Average Outlier Rank).

A detailed description of the methods included in the package can be found in Costa, E., & Papatsouma, I. (2025). A novel framework for quantifying nominal outlyingness.

Installation

The package is available on CRAN and can therefore be installed using the following:

install.packages("SONO")

The package can also be installed directly from GitHub using devtools.

# install.packages("devtools")
# devtools::install_github('EfthymiosCosta/SONO')
install.packages("SONO", repos = c('https://efthymioscosta.r-universe.dev', 'https://cloud.r-project.org'))