Unsupervised learning on multimedia data: a Cultural Heritage case study

Integrating and analyzing a large amount of data extracted from different sources can be considered a key asset for businesses, organizations, research institutions that also deal with the Cultural Heritage domain. In the last decade, Internet of Things (IoT) technologies and the massive use of mobile devices contributed to generate an enormous flow of multimedia data, whose collection, analysis and interpretation allows for real-time analysis related to the behaviours, preferences and opinions of users. In this paper we present and discuss an unsupervised learning approach on multimedia features of a dataset coming from an Internet of Things framework. The main research objective of this work is to assess how the collection of behavioural IoT data coming from the Cultural Heritage domain can be opportunely exploited by means of unsupervised learning techniques in order to produce useful insights for the stakeholders, especially considering the multimedia features of such data. The presented experimental results, executed in a real case study, assess how the Cultural Heritage domain, and the related stakeholders, can benefit from these kind of services and applications.