KUMOI, Gendo

写真a

Affiliation

Faculty of Science and Engineering, School of Creative Science and Engineering

Job title

Research Associate

Research Experience 【 display / non-display

  • 2019.04
    -
    Now

    早稲田大学・理工学術院・創造理工学部   経営システム工学科   助手

Professional Memberships 【 display / non-display

  •  
     
     

    IEEE

  •  
     
     

    THE JAPAN SOCIETY FOR MANAGEMENT INFORMATION

  •  
     
     

    the Meteorological Society of Japan

  •  
     
     

    情報処理学会

 

Research Areas 【 display / non-display

  • Chinese philosophy, Indian philosophy and Buddhist philosophy

  • Chinese philosophy, Indian philosophy and Buddhist philosophy

  • Atmospheric and hydrospheric sciences

  • Childhood and nursery/pre-school education

  • Safety engineering

display all >>

Papers 【 display / non-display

  • Feature Transfer Based Clustering for Designing Customers Growth Measures

    Yosuke Hirano, Tianxiang Yang, Gendo Kumoi, Haruka Abe, Tetsuya Tachibana, Masayuki Goto

      62 ( 10 ) 1704 - 1715  2021.10  [Refereed]

  • Deep Learning with Data Augmentation to Add Data Around Classification Boundaries

    Hideki Fujinami, Gendo Kumoi, Masayuki Goto

    Industrial Engineering & Management Systems   20 ( 3 ) 384 - 397  2021.09  [Refereed]

    DOI

  • A Study on the Optimization of the ECOC Method for Multi-label Classification Problems

      14 ( 3 ) 1 - 10  2021.08  [Refereed]

    Authorship:Lead author

     View Summary

    One of the methods for constructing a multi-valued classifier that uses a combination of given two-valued classifiers is the Error-Correcting Output Coding (ECOC) method, which is based on error-correcting codes introducing a code theory framework. Although it is experimentally known that this method performs well on real data, the theoretical optimality of the classification accuracy for the ECOC method has not been clarified. In this study, we show sufficient conditions for the ECOC method to be an optimal multi-valued classification method under the assumption that binary classifiers achieve maximum posterior probability classification. As a result, we can show that n-vs-all and Exhaustive signs are the best multi-valued classification method under the same assumptions. This suggests one of the directions of the optimization debate for various ECOC methods.

    CiNii

  • A Model on Answering Documents Retrieval Considering Diversity Based on Topic Models

    OKAWA Junya, KUMOI Gendo, GOTO Masayuki

    Journal of the Japan Society for Management Information   30 ( 1 ) 31 - 46  2021.06  [Refereed]

     View Summary

    <p>A general QA system realizes automatic answering by analyzing questions given by the user based on a model that retrieves appropriate answer candidates for each question from a set of documents (an answering documents retrieval model). When constructing an answering documents retrieval model for question/answer documents for a community QA (Question Answering) sites (cQA sites), a method based on a similarity measure on question documents can be a basic method. However, it is difficult to present appropriate answering documents while accurately grasping the diversity of answer documents that exist for each question as seen on a cQA site. In this study, we propose a method for constructing an answering documents retrieval model that can consider the diversity of answer documents by using a topic model. In order to verify the effectiveness of a proposed model, a verification experiment using question/answer documents actually posted on a cQA site is performed.</p>

    DOI CiNii

  • A Latent Class Analysis for Item Demand Based on Temperature Difference and Store Characteristics

    Yuto Seko, Ryotaro Shimizu, Gendo Kumoi, Tomohiro Yoshikai, Masayuki Goto

    Industrial Engineering & Management Systems   20 ( 1 ) 35 - 47  2021.03  [Refereed]

    DOI

display all >>

Books and Other Publications 【 display / non-display

  • 本願寺白熱教室 お坊さんは社会で何をするのか?

    雲居 玄道( Part: Joint author, ウェブに見る宗教の公共性)

    法蔵館  2015.06

Misc 【 display / non-display

  • Time Window Topic Model for Analyzing Customer Browsing Behavior

    Fumiyo Ito, Gendo Kumoi, Masayuki Goto

    IEEE 12th International Workshop on Computational Intelligence and Applications    2021.11  [Refereed]

  • A Study on New Product Recommendation Using Multi-Label CVAE for Fresh Flowers

    Aya Kitasato, Gendo Kumoi, Masayuki Goto

    IEEE 12th International Workshop on Computational Intelligence and Applications    2021.11  [Refereed]

  • Store Analysis Using Latent Representation of Robust Variational Autoencoder Based on Sales History Data

    Ryogo Okubo, Ryosuke Uehara, Gendo Kumoi, Masayuki Goto, Tomohiro Yoshikai

    The 19th Asian Network for Quality Congress    2021.10  [Refereed]

  • A Node Sharing Learning Method for Deep Neural Networks in Multi-Label Classification

    Kodai Ishikura, Aya Kitasato, Gendo Kumoi, Masayuki Goto

    The 19th Asian Network for Quality Congress    2021.10  [Refereed]

  • A Discussion on Improving Fraud Detection Performance by Generative Adversarial Networks for Transactions Data

    Guanyu Yang, Yuki Tsuboi, Ryotaro Shimizu, Gendo Kumoi, Masayuki Goto

    The 19th Asian Network for Quality Congress    2021.10  [Refereed]

display all >>

Awards 【 display / non-display

  • 2021年度論文賞

    2021.11   経営情報学会   潜在的ディリクレ配分法を用いた問合せ文書と回答文書の関係分析モデルとその応用に関する一考察

    Winner: 大川 順也, 雲居 玄道, 後藤 正幸

  • 企業賞,DBSJ特別賞

    2020.11   国立情報学研究所 情報学研究データリポジトリIDRユーザフォーラム   Knowledge Graph Attention Networkに基づく購買行動分析モデルに関する一考察

    Winner: 伊藤史世, 張志穎, 雲居玄道, 後藤正幸

  • The World CIST'18 best paper award

    2018.03   6th World Conference on Information Systems and Technologies (WorldCist'18)  

    Winner: Shigeichi Hirasawa, Gendo Kumoi, Manabu Kobayashi, Masayuki Goto, Hiroshige Inazumi

Presentations 【 display / non-display

  • A Study on the Construction Method of Shared Structure for Deep Neural Network in Multi-Label Classification

    ISHIKURA Kodai, KITASATO Aya, KUMOI Gendo, GOTO Masayuki

    Proceedings of the Annual Conference of JSAI  The Japanese Society for Artificial Intelligence

    Presentation date: 2021.06

    Event date:
    2021.06
     
     

     View Summary

    <p>Recently, the importance of techniques related to multi-label classification, which assumes that multiple labels are assigned to a single document, has been increasing. One of the approaches to solve this problem is Branched Multi-Task Networks (BMTN), which constructs a network in which the middle layer of the Deep Neural Network is shared by labels that are highly related. In BMTN, the shared structure is determined by clustering the similarity between the labels, but the number of clusters in each middle layer must be set in advance by the analyst. Therefore, it doesn't adequately represent the relationship between labels. In this study, we propose an algorithm for determining the number of clusters that can adequately represent the relationship between labels in clustering. Finally, we apply the proposed method to the Yomiuri article data, and show the usefulness of the proposed method in terms of estimation accuracy.</p>

  • Store Analysis Using Latent Representation of Robust Variational Autoencoder Based on Showing History Data

    OKUBO Ryogo, UEHARA Ryosuke, KUMOI Gendo, GOTO Masayuki, YOSHIKAI Tomohiro

    Proceedings of the Annual Conference of JSAI  The Japanese Society for Artificial Intelligence

    Presentation date: 2021.06

    Event date:
    2021.06
     
     

     View Summary

    <p>In this research, we propose a model to analyze the characteristics among stores focusing on prepared food products to eliminate food loss, targeting a retail chain with multiple stores. The data is sparse in that the number of prepared food items displayed in each store is only about 10\% of the total number of prepared food items sold in all stores. In order to cope with this sparsity, it is difficult to apply a simple compression method because of the large variation in the input data due to the existence of stores that sell unique products. The latent representation of RVAE is output as a probability distribution, and in general, the similarity is measured by sampling from this probability distribution. The latent representation of RVAE is output as a probability distribution and similarity is measured by sampling from this probability distribution. In this research, we propose a method to calculate the distance between probability distributions without sampling. We can detect stores with similar tendencies. In addition, the reconstruction error obtained by RVAE enables us to detect stores whose tendency is significantly different from other stores. Finally, we apply the proposed method to real data and verify its effectiveness.</p>

  • 仮想通貨取引データに対する敵対的生成ネットワークを用いた分類性能向上手法の検討

    楊 冠宇, 清水良太郎, 雲居玄道, 後藤正幸

    日本経営工学会春季大会 

    Presentation date: 2021.05

    Event date:
    2021.05
     
     
  • 深層異常検知モデルの中間表現によるデータ分析手法に関する一考察

    北里 礼, 相木将寛, 雲居玄道, 後藤正幸

    日本経営工学会春季大会 

    Presentation date: 2021.05

    Event date:
    2021.05
     
     
  • Robust Variational Autoencoderの潜在表現による店舗分析モデルに関する一考察

    大久保亮吾, 上原諒介, 雲居玄道, 後藤正幸, 吉開朋弘

    日本経営工学会春季大会 

    Presentation date: 2021.05

    Event date:
    2021.05
     
     

display all >>

Specific Research 【 display / non-display

  • 予報現業支援のための気象情報に対する仮説発見および天気予報文自動生成

    2020  

     View Summary

    気象予報士が気象状況を読み解く力には経験により差が生じるという問題がある.そこで本研究では画像認識で用いられるAttention Branch Networkに対し数値データである全球数値予報モデルを入力として,入力ごとにマルチタグでタグを推定するとともに,タグごとにAttention Mapを出力することによりタグごとの注視領域を特定する方法を提案する.GSMを用いてタグを推定することで,予想天気図より高精度な気象状況の推定が可能となると考えられる.さらに,Attention Mapを活用することにより,各物理量の組み合わせにより推定されるタグの対象地域の可視化が可能になると考えられる.

  • 高次元時系列データに対する仮説発見のためのパターン認識手法

    2020  

     View Summary

    近年,ネットワークに接続された電子機器から,その利用履歴である多次元時系列データが収集可能となっている.電子機器メーカーの恒常的なマーケティング活動において,顧客であるユーザの電子機器の利用履歴データを詳細に分析することは多大な価値を生むことが期待できる.しかし,取得された膨大なデータを全て人手で分析することは,計算量や人的コストの点から現実的ではない.そこで本研究では,対象となる電子機器の利用における消耗品のの選択要因に注目する.その上で,遺伝的アルゴリズムに基づく特徴量選択法を提案する.これにより恒常的な運用が可能な特徴量を同定が可能になることを示す.

  • 高次元時系列データに対する次世代パターン認識手法の開発と応用による仮説発見

    2019  

     View Summary

    近年,ユビキタスセンシングシステムに代表されように,機器は,数百から数万ものセンサが実装されている.そして,これらのセンサから時々刻々と計測されるデータが,IoT技術の発展に伴い,データを収集・蓄積することが可能となっている.そのため,これら高次元時系列データに対しての研究が活発化している.これらのログデータは高次元時系列データであり,従来技術を単純に適用できない複数の問題がある.1つは,データの獲得コストやモデル複雑さ,解釈性の困難さなどである.そこで,本研究では,マーケティング施策立案のための仮説発見を目的とし手法の提案を行った.