Intensional query answering : a clinical case study

In the current years, the evolution of informatics and the availability of a huge amount of data have raised the problem of dealing effectively and efficiently with the so-called Big Data challenge. The problem of maintaining, understanding and analyzing these data has stimulated many researches on methods and techniques to find significant correlations or associations among them, thus providing summarized views of the dataset. One of the most efficient approaches is providing implicit (intensional and approximate) knowledge by using data mining techniques, with the aim of storing association rules as a synthetic description of the data. Such information can be returned as intensional answers to the user queries, providing a synthetic, albeit approximate information on dataset contents. Indeed, inexperienced users need the support of knowledge-retrieval systems that enable them to search, retrieve and “highlight” quick, though approximate information starting from simple inputs. This work has as aim the creation of a system for intensional query-answering, “IQ4GenData”. This application allows the extraction and storage of intensional knowledge- in the form of interesting association rules - from a clinical dataset, and provides a template for writing queries over the previously mined association rules and its GUI guides the user in composing queries to be applied to this intensional, approximate knowledge. The present work has been based on the following step: • Understanding the application domain to have a clear vision on the domain of interest, by studying the clinical data that will lead to the extraction and selection Data that will be used by the application, namely the data to which we want to apply the techniques of mining. • Performing discretization on the target data in order to enrich the results of the data mining algorithm in extracting the association rules by using Weka software or as an alternative performing queries. • Investigating the system by extracting different sets of association rules according to different initial values of support and confidence and studying the mined association rules by performing queries on the intensional database.