This paper presents a new analysis support system for analyzing non-dominated solutions (NDSs) derived by evolutionary multi-criterion optimization (EMO). The main features of the proposed system are to use association rule analysis and to perform a multi-granularity analysis based on a hierarchical tree of NDSs. The proposed system applies association rule analysis to the whole NDSs and derives association rules related to NDSs. And a hierarchical tree is created through our original association rule grouping that guarantees to keep at least one common features. Each node of a hierarchical tree corresponds to one group consisting of association rules and is fixed in position according to inclusion relations between groups. Since each group has some kinds of common features, the designer can analyze each node with previous knowledge of these common features. To investigate the characteristics and effectiveness of the proposed system, the proposed system is applied to the concept design problem of hybrid rocket engine (HRE) which has two objectives and six variable parameters. HRE separately stores two different types of thrust propellant unlike in the case of usual other rockets and the concept design problem of HRE has been provided by JAXA. The results of this application provided possible to analyze the trends and specifics contained in NDSs in an organized way unlike analysis approaches targeted at the whole NDSs.

抄録全体を表示

We present a method for detecting community structures based on centrality value and node closeness. Many real world networks possess a scale-free property. This property makes community detection difficult especially on the widely used algorithms that are based on modularity optimization. However, in our algorithm, communities are formed from hub nodes. Thus communities with scale-free property can be identified correctly. The method does not contain any random element, nor requires pre-determined number of communities. Our experiments showed that our algorithm is better than algorithms based on modularity optimization in both real world and computer generated scale-free datasets.

抄録全体を表示

With the growth of Social Bookmark Services, such as Delicious, Digg or Hatena Bookmark, there is a large amount of data which can be expressed by tripartite network. Analyzing these tripartite networks is important, and community extraction is the method often used for this analysis. In this paper, we study the problems of extracting communities from tripartite networks based on modularity. Modularity is a measure to evaluate network partitioning, and there are several tripartite modularities proposed as the extension of Newman modularity. We identify the problems of these conventional tripartite modularities in network partitioning evaluation, especially when noisy edges are included in the network, and propose two new tripartite modularities that provide solutions to the problems. The results of the synthetic tripartite network experiments help us verify that our proposed methods evaluate network partitioning more effectively than the conventional ones. We also confirm meaningful community extraction results on a small real-world tripartite network. Furthermore, we propose a method for clustering edges as a means to explore network partitioning with large modularity values. It yields better community extraction results on synthetic tripartite network experiments compared with the method of clustering nodes directly.

抄録全体を表示

In this paper, we describe a novel GUI for human constraints selection in interactive constrained clustering. Clustering is one of the most popular data mining technologies, and in particular, interactive constrained clustering that uses constraints obtained from a human is promising for practical applications. However, some constraints are not effective in constrained clustering, and the cognitive load necessary for a user to give constraints is high in interactive clustering, thus we need to provide a mechanism in which a user can easily select only effective constraints for an interactive constrained clustering system. Since the constraints are considered to be training data for classification learning, traditional computational active learning like uncertainty sampling might be useful in non-interactive constrained clustering. For interactive constrained clustering, we should use human constraints selection. Thus we build a GUI for human constraints selection in interactive constrained clustering. Our GUI has two functions to effectively derive human constraints selection; exposing the effects of given constraints and providing multiple viewpoints. As the first function, we propose a GUI that can expose the effects of given constraints by emphasizing them at the clustered results. The second function can provide multiple viewpoints that a user can flexibly change to derive human constraints selection. We fully implemented an interactive constrained clustering system with the proposed GUI. We also conducted an evaluation experiment on image clustering with participants, and obtained results to support the effectiveness of our approach to derive human constraints selection.

抄録全体を表示

This paper proposes a fast word lattice generation algorithm for Japanese morphological analysis. We conducted experiments on three Japanese data sets to demonstrate that the previously proposed pruning-based algorithm is in fact not efficient enough, and that the pipeline algorithm, which is introduced in this paper, achieves considerable speed-up without loss of accuracy. Moreover, the compactness of the lattice generated by the pipeline algorithm was investigated from both theoretical and empirical perspectives.

抄録全体を表示