Fuzzy c-means (FCM) approach is the method for partitioning data into clusters by minimizing an objective function. Therefore, it is important to devise the objective function from which a simple clustering algorithm can be derived. An entropy term was introduced by Miyamoto in the FCM objective function. We proposed an objective function of the fuzzy counterpart of Gaussian mixture models (GMMs) clustering. The objective function is based on Kullback-Leibler divergence instead of the entropy. Miyamoto derived a hard clustering algorithm by linearizing the Kullback-Leibler divergence term of the objective function. In the hard c-means (HCM) clustering approach, covariance matrices are introduced as decision variables. Taking into account this method, for quick and stable convergence of FCM-like clustering, we propose the semi-hard clustering approach by constraining the membership in an interval. The semi-hard clustering result is used for a classifier design. The novel membership function suggested by GMMs and our generalized FCM approach is used for the classifier. The free parameters of the membership function are selected by particle swarm optimization (PSO). In terms of classification performance on UCI benchmark data, the classifier is comparable to the support vector machine (SVM) and surpasses the k-nearest neighbor (k-NN) classifier.
In relation to workshops, which are events designed for participatory learning and creative endeavors in groups, repeated reflections upon the activities on the workshop are important for participants and organizers. We propose a reflection-assistance method called “Workshop Reflector”, which is a seamlessly combined system of “Timeline Reflector” and “Card Reflector.” Timeline Reflector can support workshop reflection with timeline-based representation. Card Reflector can facilitate workshop reflection in a different manner than that of a conventional timeline-based representation. The cards are segmented by users from the timeline contents as noticeable events. We examined the combined use of Card Reflector and Timeline Reflector by developing a prototype.
Recently, the number of the web sites such as Web user reviews and web logs where users can express their private ideas and opinions is increasing. However, it is very difficult to read whole the reviews on the internet. The conventional systems for analysis of reviews can extract evaluation information, but they are not enough to analyze time-series variation of them. This study tries to develop a review analysis system which shows the relationship among user reviews by MDS and shows evaluation information for products using HK Graph which can visualize the relationship between words with hierarchical network structure based on the co-occurrence information for the keyword graph. This paper proposes the visualization method to analyze time-series variation of Web user reviews based on the similarity of extracted evaluation keywords.
This paper is concerned with output feedback control design for a fuzzy system with immeasurable premise variables. It is well known that Takagi-Sugeno fuzzy model describes a wide class of nonlinear systems especially when its premise variables include immeasurable functions. However, when it comes to control design of such a fuzzy system with immeasurable premise variables, a conventional parallel distributed compensator (PDC) is not feasible because it shares the same premise variables as those of a fuzzy system. In this paper, we introduce an output feedback controller with the estimate of the premise variables of an original fuzzy system. We then formulate the stabilization problem for a fuzzy system with immeasurable premise variables. Our control design method is based on a set of strict LMI conditions. No tuning parameter is necessary a priori to solve LMI conditions. Our method includes tuning matrices for control gains in a controller and hence they can be chosen to optimize the control performance of the system. Numerical examples are finally given to illustrate our control design method.
In this paper, we focus on hierarchical multiobjective linear programming problems where multiple decision makers in a hierarchical organization have their own multiple objective linear functions together with common linear constraints, and propose an interactive fuzzy decision making method to obtain the satisfactory solution which reflects not only the hierarchical relationships between multiple decision makers but also their own preferences for their membership functions. In the proposed method, instead of Pareto optimal concept, the generalized Λ-extreme point concept is introduced, which is defined in membership space. In order to obtain the satisfactory solution from among a generalized Λ-extreme point set, an interactive algorithm based on linear programming is proposed, and interactive processes are demonstrated by means of an illustrative numerical example.