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 35, Issue 2
Displaying 1-13 of 13 articles from this issue
Show all abstracts
    • |<
    • <
    • 1
    • >
    • >|
Original Papers
  • MAPPING METHOD OF PHARMACEUTICAL COMPOUNDS BY TOPOLOGICAL DATA ANALYSIS BASED ON COSA DISSIMILARITIES
    Yoshitake Kitanishi, Fumio Ishioka, Masaya Iizuka, Koji Kurihara
    2022 Volume 35 Issue 2 Pages 49-67
    Published: 2022
    Released on J-STAGE: June 01, 2023
    DOIhttps://doi.org/10.20551/jscswabun.35.2_49
    JOURNAL FREE ACCESS
    Show abstractHide abstract
      Recent years have witnessed the accumulation of vast amounts of complicated data and information. Classification and visualisation of these data are important as the first step of analysis. However, in the conventional general clustering method, all attribute information is handled equally, resulting in noise and obscuring the true structure. Another issue is how to spatially capture the characteristics of the data and robustly visualise the update and increase of the data. To solve these problems, this paper proposes the combination method of Clustering Objects on Subsets of Attributes (COSA) which captures attribute information as a subset and calculates a distance matrix, and a topological data analysis mapper (TDA Mapper) that visualises complex data structures as shapes. Furthermore, we confirm its effectiveness with extended data based on the iris data, and an application example for mapping drug data is shown.
    View full abstract
    Download PDF (10380K)
  • CHARACTERISTICS OF CUSTOMERS DELAYING LOAN REPAYMENT IN DEVELOPING COUNTRY MOTORCYCLE MARKETS
    Ryota Fujinuma, Yumi Asahi
    2022 Volume 35 Issue 2 Pages 69-85
    Published: 2022
    Released on J-STAGE: June 01, 2023
    DOIhttps://doi.org/10.20551/jscswabun.35.2_69
    JOURNAL FREE ACCESS
    Show abstractHide abstract
      Country X, a developing country in Latin America, has a large number of citizens who are late in repaying their loans. Customers who are late on their loans are very likely to default, and companies are unable to recover their manufacturing costs. This paper analyzes loan data for motorcycle sales in country X. Based on models using logistic regression, Random Forest, and XGBoost, we propose a credit risk model that can appropriately identify customers who are late in repaying their loans. In addition, logistic regression analysis was used to identify the characteristics of customers who are late in repaying their loans. Factors that influence customers who delay loan repayment are customers with low credit scores of financial institutions and customers who take out loans beyond their ability to repay, both for the rich and the poor. Factors influencing customers who did not delay loan repayment were customers' income stability and high income. The prediction accuracy of the Random Forest model was found to be the most accurate.
    View full abstract
    Download PDF (437K)
Reviews
  • A RECOMMENDER SYSTEM FOR PROMOTING USE OF HOUSEKEEPING SERVICES BASED ON ENSEMBLE NMF
    Ippei Nagai, Tomoaki Iwasaki, Fumiaki Saitoh
    2022 Volume 35 Issue 2 Pages 87-97
    Published: 2022
    Released on J-STAGE: June 01, 2023
    DOIhttps://doi.org/10.20551/jscswabun.35.2_87
    JOURNAL FREE ACCESS
    Show abstractHide abstract
      As women's social advancement has progressed in recent years, housekeeping services have attracted attention and are beginning to become widely recognized. However, the actual utilization rate is low, and the continuous utilization rate is low. In this study, we aim to improve the profit margin based on questionnaire data related to customer satisfaction and try to build a service recommendation system aimed at continuous service use and expanding new service use. By focusing on the initial value dependence of non-negative matrix factorization, which is widely used in recommendation systems, and extending this to the ensemble model, we constructed a system with stable recommendation content. Furthermore, based on the learning results of non-negative matrix factorization, we attempted an analysis that linked the recommendation content and customer image.
    View full abstract
    Download PDF (959K)
Report of Activities
Editorial Board
    • |<
    • <
    • 1
    • >
    • >|
feedback
 Top