Symbolic data analysis (SDA) represents an innovative field of research aimed at the analysis of non-punctual data. SDA defines statistical units as concepts (i.e. symbolic objects) with the possibility of assuming, for each variable, multi-values (e.g. intervals, frequency distributions, multiple categories). They can arise by synthesis of the observed data or can be defined based on expert knowledge of the domain. Observations are then represented in the form of Symbolic Data according to these new data structures.
SDA has become a flourishing and recognising scientific domain for its ability to handle large data sets through a suitable way to model their summaries. Symbolic data analysis methods developed over the past two decades allow data to be analysed in aggregate form using generalised statistical and machine learning techniques. The purpose of this IPS is to enhance the methodological and applicative contributions of SDA in the field of Data Science.
The purpose of this IPS is to enhance the methodological and applicative contributions of SDA in the field of Data Science. The session will highlight how SDA provides appropriate analytical and, above all, explanatory tools in the handling of complex, aggregated and structured data, typical of today's new challenges.
Innovative contributions will then be presented in the treatment of data in the form of intervals, through models and clustering techniques generalised to this type of data, as well as of data in the form of distributions, proposing new similarity measures applied to the context of sentiment analysis. These are intended to open up a new and more intensive development of SDA in the context of AI & ML explainability with data analysis tools applied to complex and structured data.