Using Formal Concept Analysis for Mining and Interpreting Patient Flows within a Healthcare Network
This paper presents an original experiment based on frequent itemset search and lattice based classification. This work focuses on the ability of iceberg-lattices to discover and represent flows of patient within a healthcare network. We give examples of
- PDF / 719,253 Bytes
- 6 Pages / 430 x 660 pts Page_size
- 93 Downloads / 179 Views
1 ´ Equipe Orpailleur, LORIA, Vandoeuvre-l`es-Nancy, France Laboratoire SPI-EAO, Facult´e de M´edecine, Vandoeuvre-l`es-Nancy, France
Abstract. This paper presents an original experiment based on frequent itemset search and lattice based classification. This work focuses on the ability of iceberg-lattices to discover and represent flows of patient within a healthcare network. We give examples of analysis of real medical data showing how Formal Concept Analysis techniques can be helpful in the interpretation step of the knowledge discovery in databases process. This combined approach has been successfully used to assist public health managers in designing healthcare networks and planning medical resources. Keywords: Formal Concept Analysis, frequent itemsets, network.
1
Introduction
Knowledge Discovery in Databases (KDD) is an iterative and interactive process for identifying valid, novel, and potentially useful patterns in data [1]. KDD is usually divided into three main steps: data preparation, data mining, and interpretation of the extracted units. Data mining, often considered as the central step in this process, is still an active field of research. The success key in KDD practice relies also on ability of easily producing units understandable as knowledge units. One way of achieving such a goal relies on an adapted visualization of the extracted units. In this paper, we present an original experiment based on both frequent itemset search and lattice-based classification. This experiment holds on medical data and is aimed at showing the interactions and collaborations between hospitals in the French Region of Lorraine. This experiment may be regarded from two points of view: on the one hand, it is based on frequent itemset search on a medico-economic database, and on the other hand, the visualization of extracted units is based on Formal Concept Analysis (FCA) techniques [2], organizing the extracted units into a lattice for medical analysis and interpretation. At our knowledge, this is an original combination of data mining and FCA techniques that has been rarely carried on until now. Indeed, this is one of the main feature of this paper to show how FCA techniques can be very helpful in the interpretation step of KDD process. The results of this experiment have been used by healthcare administration in Lorraine for planning and evaluation purposes [3]. S. Ben Yahia et al. (Eds.): CLA 2006, LNAI 4923, pp. 263–268, 2008. c Springer-Verlag Berlin Heidelberg 2008
264
2
N. Jay, F. Kohler, and A. Napoli
Health Networks and Collaborations
Healthcare networks are sets of healthcare actors working in cooperation, sharing information, and providing care for the same patients. In France, some networks are formally structured but others are still in an implicit existence. Thus, healthcare policy should be based on this current state of things to plan new networks or optimize existing ones. However, for both structured and implicit networks, knowledge on the degree of collaboration between hospitals is poor, because
Data Loading...
