Data Mining Techniques for Complex User-Generated Data

Nowadays, the amount of collected information is continuously growing in a variety of different domains. Data mining techniques are powerful instruments to effectively analyze these large data collections and extract hidden and useful knowledge. Vast amount of User-Generated Data (UGD) is being created every day, such as user behavior, user-generated content, user exploitation of available services and user mobility in different domains. Some common critical issues arise for the UGD analysis process such as the large dataset cardinality and dimensionality, the variable data distribution and inherent sparseness, and the heterogeneous data to model the different facets of the targeted domain. Consequently, the extraction of useful knowledge from such data collections is a challenging task, and proper data mining solutions should be devised for the problem under analysis. In this thesis work, we focus on the design and development of innovative solutions to support data mining activities over User-Generated Data characterised by different critical issues, via the integration of different data mining techniques in a unified frame- work. Real datasets coming from three example domains characterized by the above critical issues are considered as reference cases, i.e., health care, social network, and ur- ban environment domains. Experimental results show the effectiveness of the proposed approaches to discover useful knowledge from different domains.