Creat membership Creat membership
Sign in

Forgot password?

Confirm
  • Forgot password?
    Sign Up
  • Confirm
    Sign In
home > search

Now showing items 65 - 80 of 85

  • Trajectory mining using multiscale matching and clustering

    Shusaku Tsumoto   Shoji Hirano  

    This paper focuses on clustering of trajectories of temporal sequences of two laboratory examinations. First, we map a set of time series containing different types of laboratory tests into directed trajectories representing temporal change in patients’; status. Then the trajectories for individual patients are compared in multiscale and grouped into similar cases by using clustering methods. Experimental results on the chronic hepatitis data demonstrated that the method could find the groups of trajectories which reflects temporal covariance of platelet, albumin and choline esterase.
    Download Collect
  • Mining Complex Data

    Zbigniew W. Ras   Shusaku Tsumoto   Djamel A. Zighed  

    In recent years; the complexity of data objects in data mining applications has increased as well as their plain numbers. As a result; there exist various feature transformations and thus multiple object representations. For example; an image can be described by a text annotation; a color histogram and some texture features. To cluster thesemulti-represented objects; dedicated datamining algorithms have been shown to achieve improved results. In this paper; we will therefore introduce a method for hierarchical density-based clustering of multi-represented objects which is insensitive w.r.t. the choice of parameters. Furthermore; we will introduce a theoretical model that allows us to draw conclusions about the interaction of representations. Additionally; we will show how these conclusions can be used for defining a suitable combination method for multiple representations. To back up the usability of our proposed method; we present encouraging results for clustering a real world image data set that is described by 4 different representations.
    Download Collect
  • Mining Temporal Patterns in Time-series Medical Databases: A Hybrid Approach of Multiscale Matching and Rough Clustering

    Shoji Hirano   Xiaoguang Sun   Shusaku Tsumoto  

    This paper presents a method for analyzing time-series laboratory examination databases. The key concept of this method is classification of temporal patterns using multiscale structure matching and a rough set-based clustering method. Multiscale matching enables us to capture similarity between two sequences of examinations from both short-term and long-term points of view. The rough-set based clustering technique is then applied to classify the sequences according to the relative similarity obtained through multiscale matching. In the experiments we show that this hybrid approach can be used to discover interesting temporal patterns hidden in the time-series databases. keywords: rough sets; multiscale strcture matching; time-series analysis.
    Download Collect
  • 330 Granularity and statistical independence in a contingency table

    Shusaku Tsumoto  

    " This paper gives a relations between the degree of granularity and that of dependence of contingency tables. From the results of determinantal divisors; it seems that the devisors provide information on the degree of dependencies between the matrix of the whole elements and its submatrices and the increase of the degree of granularity may lead to that of dependence. However; this paper shows that a constraint on the sample size of a contingency table is very strong; which leads to the evaluation formula where the increase of degree of granularity gives the decrease of dependency."
    Download Collect
  • Comparison of induced rules based on likelihood estimation

    Shusaku Tsumoto   Shimane Medical Univ.   Izumo Shimane   Japan.  

    Rule induction methods have been applied to knowledge discovery in databases and data mining; The empirical results obtained show that they are very powerful and that important knowledge has been extracted from datasets. However; comparison and evaluation of rules are based not on statistical evidence but on rather naive indices; such as conditional probabilities and functions of conditional probabilities. In this paper; we introduce two approaches to induced statistical comparison of induced rules. For the statistical evaluation; likelihood ratio test and Fisher's exact test play an
    Download Collect
  • Computational Intelligence in Automotive Applications

    Ying Liu   Aixin Sun   Han Tong Loh   Wen Feng Lu   Tsau Young Lin   Ying Xie   Anita Wasilewska   Ivo Dolezel   Tomasz G. Smolinski   Mariofanna G. Milanova   Shuichi Iwata   Yukio Ohsawa   Shusaku Tsumoto   Ning Zhong   Yong Shi   Einoshin Suzuki   Fabrice Guillet   Lakhmi C. Jain   Danil V. Prokhorov  

    and reference the set of actual center-of-lane paths (Lane Segments). The Mobility control module managesthis relationship by constantly updating the list of Lane Segments for a particular Lane Element based on real-timesensing of the lane position and curvature Intelligent Control of Mobility Systems 249 Right Turnat3-wayIntersectionOwn Vehicle’s Path
    Download Collect
  • Automated knowledge acquisition from clinical databases based on rough sets and attribute-oriented generalization

    Shusaku Tsumoto  

    Rule induction methods have been proposed in order to acquire knowledge automatically from databases. However, conventional approaches do not focus on the implementation of induced results into an expert system. In this paper, the author focuses not only on rule induction but also on its evaluation and presents a systematic approach from the former to the latter as follows. First, a rule induction system based on rough sets and attribute-oriented generalization is introduced and was applied to a database of congenital malformation to extract diagnostic rules. Then, by the use of the induced knowledge, an expert system which makes a differential diagnosis on congenital disorders is developed. Finally, this expert system was evaluated in an outpatient clinic, the results of which show that the system performs as well as a medical expert.
    Download Collect
  • Hierarchical clustering of asymmetric proximity data based on the indiscernibility-level

    Shusaku Tsumoto   Shoji Hirano  

    In this paper; we present a method for clustering asymmetric proximity data. First; we calculate the indiscernibility level for each object pair; that quantifies the level of global agreement for regarding the two objects as indiscernible. Then; hierarchical linkage grouping is applied to unite objects according to the derived indiscernibility level. This scheme enables users to examine the hierarchy of data granularity and obtain the set of indiscernible objects that meets the given level of granularity. Additionally; since indiscernibility level is derived based on the binary classifications determined independently for each object; it can be applied to non-Euclidean; asymmetric relational data. Using a synthetic numerical data and a real-world data about inter-prefectural movement of university students; we demonstrate that the method could represent hierarchy of data granularity and could obtain interesting groups of objects from asymmetric proximity data.
    Download Collect
  • Fuzziness from attribute generalization in information table

    Shoji Hirano   Shusaku Tsumoto  

    This paper shows some problems with combination of rule induction and attribute-oriented generalization; where if a given hierarchy includes inconsistencies; then application of hierarchical knowledge generates inconsistent rules; due to generation of fuzziness. Then; we propose an approach to solving this problem by using fuzzy linguistic variables. Also; this approach suggests that combination of rule induction and attribute-oriented generalization can be used to validate concept hiearchy.
    Download Collect
  • Statistical extension of rough set rule induction

    Shusaku Tsumoto   Shimane Medical Univ.   Izumo Shimane   Japan.  

    Rough set based rule induction methods have been applied to knowledge discovery in databases. The empirical results obtained show that they are very powerful and that some important knowledge has been extracted from datasets. However; quantitative evaluation of induced rules are based not on statistical evidence but on rather naive indices; such as conditional probabilities and functions of conditional probabilities. In this paper; we introduce a new approach to induced rules for quantitative evaluation; which can be viewed as a statistical extension of rough set methods. For this extension; chi-square distribution and F- distribution play an important role in statistical evaluation.
    Download Collect
  • Classifiers from Granulated Data Sets: Concept Dependent and Layered Granulation

    Piotr Synak   Jakub Wroblewski   RSKDWorkshop Chairs   Marzena Kryszkiewicz   Aijun An   Jan G. Bazan   Shoji Hirano   Jan Komorowski   Lech Polkowski   Andrzej Skowron   Guoyin Wang   Hui Wang   Piotr Artiemjew   Shusaku Tsumoto   Szymon Wilk  

    Granulation of data and the idea of a granular data set were proposed by L.A. Zadeh on the basis of the assumption that is at heart of all data mining techniques; i.e.; that given a plausible similarity measure on objects in a data set; objects which are similar in a satisfactory degree would have also similar or even equal decision (class) values. This assumption underlies reasoning by analogy; nearest neighbors methodology; case based reasoning and rough set methods as well. This assumption taken to an extreme implies that once a granular data system has been induced from a real data set; and decision/classification rules have been computed from it; these rules when applied to the original data should produce classification results close satisfactorily to classification results obtained on the real data with rules induced from the original non–granulated data. A number of tests performed borne out this hypothesis. We present in this work results of experiments with real data sets and we work with the two extensions of granulation techniques presented by us in the literature so far; i.e.; concept dependent granulation and layered granulation.
    Download Collect
  • Unsupervised Grouping of Trajectory Data on Laboratory Examinations for Finding Exacerbating Cases in Chronic Diseases

    Shoji Hirano   Shusaku Tsumoto  

    In this paper we present a method for finding exacerbating cases in chronic diseases based on the cluster analysis technique. Cluster analysis of time series hospital examination data is still a challenging task as it requires comparison of data involving temporal irregulariry and multidimensionalty. Our method first maps a set of time series containing different types of laboratory tests into a directed trajectory representing the time course of patient status. Then the trajectories for individual patients are compared in multiscale and grouped into similar cases. Experimental results on synthetic digit-stroke data showed that our method could yield low error rates (0.016±0.014 for classification and 0.118±0.057 for cluster rebuild). Results on the chronic hepatitis dataset demonstrated that the method could discover the groups of excacerbating cases based on the similarity of ALB-PLT trajectories.
    Download Collect
  • A method for detecting suspicious regions in mammograms based on multiscale image filtering and regression-line analysis

    Shusaku Tsumoto   Shoji Hirano  

    Early detection of breast cancer is an important task for preventing the loss of lives/breasts of the women. In this paper we propose a method for detecting suspicious features of breast cancers on mammograms by the combination of multiscale image filtering and regression-line analysis. Images are represented and analyzed at different scales. Calcifications are detected on the finest resolution; masses and mammary glands are detected on a more abstracted plane. After detecting mammary glands; we apply linear regression to the parts of mammary ducts; and estimates the degree of concentration by the measure of average minimal distance to the concentration point. Experimental results on the DDSM mammography images demonstrate that these approaches could contribute to the successful detection of these features.
    Download Collect
  • Discovery of exacerbating cases in chronic hepatitis based on cluster analysis of time-series platelet count data

    Shoji Hirano   Shusaku Tsumoto  

    "This paper reports the results of temporal analysis of platelet (PLT) data in chronic hepatitis dataset. First we briefly introduce a cluster analysis system for temporal data that we have developed. Second; we show the results of cluster analysis of PLT "
    Download Collect
  • Decomposition of pearson residuals of three-variables contingency cube

    Shoji Hirano   Shusaku Tsumoto  

    This paper shows the meaning of Pearson residuals when a contingency table is three dimensional. While information granules of statistical independence of two variables can be viewed as determinants of 2×;2- submatrices, those of three variables consist of several combinations of linear equations which will become odds ratio when they are equal to 0. Interstingly, the property on the symmetry of two dimensional tables is lost, and the lost of symmetry gives some meaning of Pearson residuals of three dimensional tables.
    Download Collect
  • 62410X Visualization of similarities and dissimilarities in rules using MDS

    Shusaku Tsumoto  

    " This paper proposes a visualization approach to show the similarity relations between rules based on multidimensional scaling (MDS); which assign a two-dimensional cartesian coordinate to each data point from the information about similiaries between this data and others data. First; semantic and synctatic similarities of rules are obtained after rules are induced from a datasets. Then; MDS is applied to each similarity. MDS visualizes the difference between semantic and synctatic simliarites. This method was evaluated on two medical data sets; whose experimental results show that knowledge useful for domain experts could be found."
    Download Collect
1 2 3 4 5 6

Contact

If you have any feedback, Please follow the official account to submit feedback.

Turn on your phone and scan

Submit Feedback