Kozo Mizutani

New Proposal and Accuracy Evaluation of Grey Prediction GM

Grey model (abbreviated as GM), which is based on Dengs grey theory, has been established as a prediction model. At present, it has been widely applied in many research fields to solve efficiently the predicted problems of uncertainty systems. However, this model has irrational problems concerning the calculation of derivative and background value z since the predicted accuracy of GM is unsatisfying when original data shows great randomness. In particular, the predicted accuracy falls in case of higher-order derivative or multivariate greatly. In this paper, the new calculation methods of derivative and background value z are first proposed to enhance the predicted power according to cubic spline function. The newly generated model is defined as 3spGM. To further improve predicted accuracy, Taylor approximation method is then applied to 3spGM model. We call the improved version as T-3spGM. Finally, the effectiveness of the proposed model is validated with three real cases.

On the Generalization of Grey Relational Analysis

In this paper we developed new grey relational analysis by expanding a range of treatable value. The classical GRA procedure deals with almost whitened values, reference vectors (or sequences), inspected vectors and grey relational grades. The classical GRA is called White-In-White-Out (WIWO) GRA in this paper. We developed two kinds of GRA algorithm based on the topological-based GRA, after defined several new operations of the interval grey number. On the one hand, GRA which provides the grey relational grade a whitened value finally is named Grey-In-White-Out (GIWO) GRA. On the other hand, GRA which provides the grey relational grade the interval grey number is named Grey-In-Grey-Out (GIGO) GRA. Both new GRA procedures deal with the reference vector and inspected vectors as the interval grey number. Three examples are given in this paper; the properties of the grey relational grade given by the proposals are discussed.

Decision rule extraction and reduction based on grey lattice classification

This paper proposes a decision rule of extraction and reduction that is based on grey lattice classification. This proposal method comes from joining between rough set theory and grey theory as an approximation algorithm. Grey lattice operations are defined by combining interval grey number in grey theory with interval lattice operations in interval algebra. By defining the equivalents in interval grey number, given data space is correspondent to equivalents of rough set. This proposal method classifies each data set into 3-patterns from given training samples, as existing possibility class, newly made possibility class and existing necessity class. As given examples which require only necessity class, decision rule is simplified by a reduction procedure.

A Realization Algorithm of Grey Structural Modeling with MATLAB

Grey structural modeling (GSM in short) is a new approach of system modeling methods succeeding to ISM and FSM. GSM has two procedures: estimating a hierarchy of given elements, estimating paths among given elements. The former procedure is constructed from complex equations. In this paper we developed one realization algorithm of the GSM procedure. The main problem we should solve is how to group given elements into several classes and to decide their hierarchy. We are possible to group analyzing an error matrix which is obtained from the localized grey relational grade, and we are also possible to decide their hierarchy according to the localized grey relational grade. We used the topological-based grey relational analysis. These procedures are shown as a pseudo language with several figures, and are realized by MATLAB. Several examples applied with the developed program are shown in this paper

A K-means Clustering Approach Based on Grey Theory

A lot of clustering algorithms based on grey system theory, especially based on the grey relational matrix, have been already reported, which finds out a centroid of each class by moving given objects as vectors. We developed new clustering procedure called grey K-means, which is able to handle the number of required clusters such as the hard K-means or the fuzzy c-means. Assume that the number of found clusters by the proposal is between 1 and the number of classified instances, a required threshold value is exist in [0,1]. We defined a value range of the threshold as the interval grey number, and the range is specified automatically until obtaining the required clusters. In addition a new clustering method which analyzes the grey relational matrix closely instead of moving vectors is suggested. Several well-known data sets in the classification problem are applied, and we discuss their performances and the optimal threshold value.

A Kansei Expression Method Based on the Simultaneous GM with the Kansei Map

We propose the K-Model using Multi-Agent Systems (MAS) and dynamic system as a Kansei information processing model. A dynamic system has been realized by Grey Model (GM) in grey system theory. However, the output system in the conventional model was one-dimensional function. And, it was specifically insufficient to show current human emotion. In the emotion analysis system from human voice, the discrimination accuracy remains at around 60%. This paper presents a new proposal for a multi-dimensional Kansei expression method. This method is constructed from the Kansei map and the simultaneous GM. The Kansei map is a map that includes Kansei elements on a two-dimensional plane. This method is introduced into the emotion analysis system from human voice and its evaluation experiment is carried out. Moreover, a data-preprocessing method to establish the GM newly proposed. As a result of the experiment, it is possible that the discrimination accuracy is improved about 20% compared with the conventional method.

Writer Identification for Offline Handwritten Kanji Characters Using Multiple Features

 Abstract—This paper presents a study on character features and recognizers used for writer identification of offline handwritten Kanji characters. It is shown that a combination of two global features, two local features, and majority voting as a recognizer is efficient for writer identification. We performed experiments using an offline Kanji character database containing one-hundred Kanji characters, each written by one-hundred writers, and fifty samples of each Kanji character for a given writer. The experimental results show that the identification rate is 7 points higher than the conventional method using a single feature and obtained an identification rate higher than 99% by using three character classes. Writer identification based on scanned images of handwritten characters is a useful biometric modality with applications in forensic and historical document analysis. Research on writer identification that uses online characters is widespread, but offline characters lack form for conveying dynamic information. Nevertheless, research on writer identification using offline characters has proposed many features to acquire useful information. We studied efficient character features and recognizers for writer identification of handwritten offline Kanji characters. Kanji consists of logographic Chinese characters adopted in Japanese writing. The text-dependent writer identification in our research uses a character recognition process before writer identification because, in text-dependent writer identification, a character class is assumed to be already recognized. Therefore, most of the research on text-dependent writer identification of handwritten characters has the following characteristics: employs features developed for character recognition, does not use multiple character features, and uses the local features of a character. In this paper, we propose efficient character features and a recognizer for text-dependent writer identification of a handwritten Kanji character (1).

Versatile English Learning System Using Webpages as Learning Materials

In Japan, there is a strong demand for English language instruction not only from students, but also from working adults. However, the nation has few systems that have functions for improving English skills listening, speaking, reading, and writing, are simple to operate, and inexpensive. To solve the problem, we propose a versatile English learning system that uses webpages as learning materials. The proposed system has following characteristics: it creates Japanese versions of webpages translated from English, it can determine a users English skill, it is capable of speaking and listening interactions, it does not require an installation fee, and it is easy to operate.