Yasunori Shiono, Tadaaki Kirishima, Yoshinori Ueda, and Kensei Tsuchida

Abstract:
Fuzzy graphs have been used frequently and effectively as a method for sociogram analysis. A fuzzy graph has the fundamental characteristic of being able to express a variety of relationships between nodes. The drawing of fuzzy graphs has been studied in computer-aided analysis systems with human interfaces and methods using genetic algorithms. However, computer-aided analysis systems with human interfaces do not provide for automatic drawing, while methods using genetic algorithms have the defect of requiring too much execution time for finding a locally optimum solution. To overcome these defects, we propose an algorithm for drawing intelligible and comprehensive fuzzy graphs using a partition tree. This method automatically draws the fuzzy graph with nodes arranged on the intersections of a latticed space. Since nodes are optimally arranged on the latticed intersections and put together at a nearby position in accordance with the transition of clusters according to cluster levels in the partition tree, drawing the algorithm makes fuzzy relations easier to understand through fuzzy graph representation. Moreover, fuzzy graphs can be drawn faster than by conventional methods. This paper describes the algorithm and its verification by introducing a system implementing the method for displaying fuzzy graphs. Moreover, we have carried out a case study in which a questionnaire has been administered to students, allowing us to analyze human relations quantitatively using a method based on fuzzy theory. Human relations are represented as fuzzy graphs by our algorithm and analyzed using the fuzzy graph.

Keywords: fuzzy graph, cluster analysis, graph drawing algorithm, sociometry analysis