Um modelo de interface extensível para sistemas de mineração de dados por regras de associação

Abstract

Currently, one of the main challenges of computing is the huge volume of data due to the storage facility and increasing use of technology in different contexts. The analysis of this data provides support for decisions in distinct areas. However, without efficient computational techniques it becomes unfeasible to analyze this large volume of data. Thus, data mining emerges as a promising field, since it allows for knowledge discovery from large volumes of data. Amongst the many techniques available for data mining, in this work we focus on association rules. Even though association Rules data mining systems are very popular they present users with a great challenge. These systems require users to have technical knowledge about data mining techniques in order to interact with them. In this work we propose an extensible interface model which aims at widening the use of data mining systems. To do so, the model allows for a new abstract high level interface specific to a context to be created. This new high level interface abstracts the technical knowledge required, making it easier to interact with the system. Based on this model, an extensible module that can be added on to 2nd generation data mining systems can be developed. The model considers two distinct user profiles: the experts and final users. Expert users are those who not only have knowledge of the domain, but also of the required technical concepts to interact with the system, whereas final users have domain knowledge, but not data mining technical knowledge. Expert users interact with the extensible module and create a new high level interface specific to final users context with which they can interact. The model is grounded on Semiotic Engineering theory, which perceives the interaction as designer-to-user mmunicative act. The model allows expert users to become co-authors of the message being transmitted by the systems, as they create new high level interfaces to final users. Preliminary evaluations of the model were executed and also a prototype was developed to provide indicators of the feasibility and utility of the model. The indicators pointed to the ability of the model to widen the use of the system to users who do not have data-mining technical knowledge at a low cost