Dimensions reordering for visual mining of association rules using parallel set

Abstract

Mining the interesting association rules from data and clusters presents high demands in many application fields, such as telephonic operator, social networks, marketing especially in decision-making process. Exploring the set of extracted rules from the high dimensional data and clusters presents a serious challenge of visual analytic process. This problem becomes more challenging when the data items increase. To deal with this challenge, the use of the interactive multidimensional visualisation technique is necessary. In our work, we base on parallel set visualisation technique as a tool of knowledge discovery tool to mining the set of rules from categorical data. In parallel set, the order and arrangement of dimensions have a major influence on the analytic process, therefore we need to find an effective measure, which can generate an expressive order. This last helps the user to explore and analyse the visual display of data. In this paper, we are interested in analysing data to detect the association rules between dimensions based on our application of entropy measures. These dimensions are reordered while the high number of rules with a high confidence is displayed. It allows the users to visualise easily the important rules in a short time.