Machine learning for compositional data analysis in Support of the Decision Making Process

In recent years, the development of digital technologies brings a
lot of changes in the way of operating, leading, and working processes
in companies. Accordingly, advanced technologies such as Artificial
Intelligent, Big Data, Internet of things, etc., are widely applied to
aggregate, transform, and analyze data, thereby inferring meaningful
information from the results, making important decisions. As a branch
of AI, machine learning (ML) is a method of data analysis that constitutes
analytical model-building automation. The main objectives of
ML are designing algorithms that can learn from data by themselves,
identify patterns, and adapt them without human intervention. The
goal of this chapter is to summarize the researches related to applying
ML to compositional data (CoDa), including principal component
analysis (PCA), clustering, classification, and regression. CoDa is a
special type of data, well-defined on the Simplex space. Since it carries
only relative information, the traditional methods can not be applied
directly to this type of data without adapting or transforming data
into normal form. Besides, we will introduce a transformation method
based on Dirichlet density estimation to transform CoDa into real data
and apply those transformed data in anomaly detection using Support
Vector Data Description (SVDD). A simulation example to illustrate
this method is also provided at the end of the chapter.