Bulletin of the Computational Statistics of Japan
Online ISSN : 2189-9789
Print ISSN : 0914-8930
ISSN-L : 0914-8930
Search
Advanced search
OR
Browse
Search
Volume 28, Issue 2
Displaying 1-16 of 16 articles from this issue
Show all abstracts
    • |<
    • <
    • 1
    • >
    • >|
Papers
  • PAIRWISE CONSTRAINED K-MEANS VIA PERMUTATION MATRIX
    Ryo Takahashi, Naomichi Makino
    2015 Volume 28 Issue 2 Pages 105-119
    Published: 2015
    Released on J-STAGE: May 01, 2017
    DOIhttps://doi.org/10.20551/jscswabun.28.2_105
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    In this paper, we proposed a novel algorithm for pairwise constrained k-means clustering. One of the major problems in the previous algorithms is that the calculation may be stopped when clusters satisfying the constraints cannot be found. The proposed algorithm can partition objects keeping the pairwise constraints using a permutation matrix and thus avoid the problem in the previous studies. A simulation study is performed for assessing an alternating least-squares algorithm for pairwise constrained k-means clustering. The developed algorithm and its applications are illustrated with the two real data examples.
    View full abstract
    Download PDF (3353K)
  • MULTIVARIATE ADAPTIVE REGRESSION SPLINE METHOD AND ITS APPLICATIONS
    Toshio Shimokawa, Mitsuhiro Tsuji, Masashi Goto
    2015 Volume 28 Issue 2 Pages 121-136
    Published: 2015
    Released on J-STAGE: May 01, 2017
    DOIhttps://doi.org/10.20551/jscswabun.28.2_121
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    In medical research, ordered categorical outcomes (such as seriousness, side effects, and grade of treatment) are sometimes used as response variables. Typically, the influencing factors are explored by ordered logistic regression. Recently, the tree-structured method has been extended to ordered categorical outcomes (Piccarreta, 2008; Archer, 2010), but the predictive outcomes of this approach are poor. In this paper, we newly develop a nonlinear ordered categorical regression method, named PO-MARS, which extends multivariate adaptive regression splines (Friedman, 1991). The PO-MARS method is developed on a proportional odds model framework, and model selection is based on the modified Akaike's information criteria (AIC) proposed by LeBlanc and Crowley (1999). The effectiveness of the PO-MARS method was illustrated through a practical example. In small-scale simulations, this method demonstrated higher predictive performance than existing methods.
    View full abstract
    Download PDF (2473K)
Reviews
  • [title in Japanese]
    [in Japanese]
    2015 Volume 28 Issue 2 Pages 137
    Published: 2015
    Released on J-STAGE: May 01, 2017
    DOIhttps://doi.org/10.20551/jscswabun.28.2_137
    JOURNAL FREE ACCESS
    Download PDF (1710K)
  • CLASSIFICATION OF CUSTOMERS AND ANALYSIS OF PURCHASE TENDENCY IN A GROUP-BUYING COUPON SITE
    Kazuto Igarashi, Yusuke Matsui, Hiroyuki Minami, Masahiro Mizuta
    2015 Volume 28 Issue 2 Pages 139-146
    Published: 2015
    Released on J-STAGE: May 01, 2017
    DOIhttps://doi.org/10.20551/jscswabun.28.2_139
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    In this paper, we propose a method for extracting useful information of the customer's purchase tendency from data recorded in a group-buying coupon site. To begin with, we classify the customers using cluster analysis according to categories of purchased coupon. We also show the relationship between the customer's attributes and the categories of coupons using correspondence analysis. Through comparison of the results for all customers and for the customers in each cluster, we discuss the in-depth interpretation.
    View full abstract
    Download PDF (2000K)
  • DESIGN OF RECOMMENDATION SYSTEM FOR REAL-ESTATE E-COMMERCE SITE BASED ON THE CO-OCCURRENCE RELATION ANALYSIS OF SEARCH CRITERION
    Kosuke Okusa, Kou Abe, Takaya Naito, Naoto Yamaguchi, Takuma Kada, Yos ...
    2015 Volume 28 Issue 2 Pages 147-154
    Published: 2015
    Released on J-STAGE: May 01, 2017
    DOIhttps://doi.org/10.20551/jscswabun.28.2_147
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    We present the design of recommendation system for real-estate e-commerce site based on the co-occurrence relation analysis of search criterion. A study on design of recommendation system is very important role in the field of e-commerce business. In this paper, We focus on the search criterion in the e-commerce site. Assuming that each search criterion have implicit relationship. We estimate the degree of association between each search criterions using zero-inflated beta distribution. As a result, our method shows good performance in the real-estate e-commerce site.
    View full abstract
    Download PDF (4003K)
  • VISUALIZATION OF BROWSING BEHAVIOR ON AN E-COMMERCE SITE
    Takahiko Ozaki, Shinya Takano, Yoshikazu Yamamoto
    2015 Volume 28 Issue 2 Pages 155-164
    Published: 2015
    Released on J-STAGE: May 01, 2017
    DOIhttps://doi.org/10.20551/jscswabun.28.2_155
    JOURNAL FREE ACCESS
    Show abstractHide abstract
    In this paper, we describe the design and implementation of our software for visualizing web site access data, and several analysis results for the data set of an E-commerce site that was provided for the date competition organized by the Joint Association Study Group of Management Science. Our software displays session histories by polygonal lines separated by their colors. We implemented our software by the Java language and MySQL database. Our software can find the characteristics of browsing behavior just before the purchase of products. Therefore, it enables us to see products that were compared with the purchased one. In the e-commerce data set, we find that four kinds of session histories are clearly distinguished: purchasing the target product only, purchasing another product only, purchasing both the target product and another product, and just viewing products.
    View full abstract
    Download PDF (6818K)
Report of Activities
Editorial Board
    • |<
    • <
    • 1
    • >
    • >|
feedback
 Top