Interpreting Logistical Performances using Trellis displays

Abstract

For the past two years we are working on benchmarking performance data to explore the potential value of knowledge discovery methods in databases for information access. Many of these data are still very under analysed. In particlular we are interested in exploiting machine-learning systems using an inductive supervised approach to predict a company’s logistical performance related to its past financial state. Benchmarking assessments produce a broad heterogeneous collection of data difficult in analysing and which goes beyond the classical framework in providing measurable information. So there is a need to describe more complex units or concepts. There is a need for carriers of aggregated information and new information at the same time. And above all there is a need to interpret these new carriers of information. In this paper we are extending our previous work in this field which was concentrated on building new carriers of information, to recent work that deals with visual displays reputed in discovering structure in complex data. And we summarize this research with some conclusions and future perspectives.