Symbolic and spatial data analysis: Mining complex data structures

Nowadays, researchers from different areas must face the growing complexity of the information available, which is characterised by the generally high dimensions of databases and their structure, resulting from the observation of multivariate phenomena in different state/space occasions. As a consequence, there is an increasing interest in extracting knowledge from large collections of data. Data mining, knowledge discovery in data bases, intelligent data analysis are some of the terms adopted to identify parallel streams of work aiming to support humans in extracting previously unknown, valid, potentially useful and understandable patterns in the data. Most studies in these areas have until recently focused on a relatively simple representation of data: a database relation, or a standard data table, or a set of points in a feature space. In fact, the relational model is clean and simple, and a relational table can be easily mapped into the mathematical concept of matrix. Moreover, many data analysis applications concern administrative data, which are easily represented by this model. With the advent of the “information age”, we have witnessed to a dramatic growth of applications in government, business and education, many of which are sources of various data, organised in different structures and formats. The chances that computers have provided have enlarged the meaning of “data”, have defined new sorts of problems in knowledge discovery, and have led to the development of completely new classes of models and data analysis algorithms. Object oriented databases, and, more recently, object-relational databases allow for the manipulation of data with complex structures, which then require novel methodologies of analysis.