Study at TCU


Name MIKAWA Kenta
Official Title Associate Professor
Affiliation Information Systems, Informatics
Profile My specialty is machine learning and its application to data analysis. Statistical machine learning is one of the methods for extracting rules behind data, and it includes supervised learning, unsupervised learning, and reinforcement learning and so on. Among them, I am working on build superior model and more precise parameter estimation of the model both supervised and unsupervised learning.
The former is to develop predictive models that include interactions, and gaining a distance measure that can adequately express statistical relationships between data.
And the latter, unsupervised learning, is to extract hidden relationships among data using probabilistic latent structure models.
I am investigating methodologies for better parameter estimation by using optimization problems and regularization methods for these methods.
In addition, my research is to develop a practical method to analyze real data by using the proposed method. And the proposed method can be used for practical analysis in marketing and other activities.
Research Field(Keyword & Summary)
  1. (1) Machine learning

    Machine learning is generally classified into two categories: supervised learning, which predicts new input data based on their statistical features given training data whose categories are known in advance, and unsupervised learning, which extracts their features using only data whose categories are unknown. In other words, supervised learning obtains rules representing the data from the given correct answers. In contrast, unsupervised learning finds a structure that represents the data well based on its features.

  2. (2) Distance metric learning

    The distance metric learning is to estimate a suitable metric matrix of the Mahalanobis distance from the training data. And the given metric matrix projects input data on the feature space, onto which input data can be easily classified into the correct category. The distance metric learning techniques can be generally divided into supervised and unsupervised distance metric learning mainly. The former is employed to acquire the distance metric from the training data whose category label is already known, while the latter is the method utilized to carry out clustering or dimensionality reduction from the input data whose category label is unknown.

  3. (3) Statistical relational learning using latent structure model

    Statistical relational learning is the method that extracts the factors of the relationship among data based on the relational data whose relationship is given as numerical data.
    In general, the statistical relational data analysis, only the relationship data are given, but we extend this model by assuming that the attribute information of the data is also given. Based on this assumption, to model the latent structure of the data and to propose knowledge discovery method.

Representative Papers
  1. (1) Analysis of entry behavior of students on job boards in Japan based on factorization machine considering the interaction among features, Cogent Engineering, Vol. 8, 2021, 10.
  2. (2) An Extension of Semi-supervised Boosting to Multi-valued Classification Problems, Total Quality Science, Vol. 6, No. 2, pp. 60 - 69, 2021, 2.
  3. (3) Relational analysis model of weather conditions and sales patterns based on nonnegative tensor factorization, International Journal of Production Research, Vol. 58, No. 8, pp. 2477 - 2489, 2019, 12.
  4. (4) An Analytic Model to Represent Relation between Finish Date of Job-Hunting and Time-Series Variation of Entry Tendencies, Industrial Engineering & Management Systems, Vol. 18, No. 3, pp. 292 - 304, 2019, 9.
  5. (5) Data pair selection for accurate classification based on information-theoretic metric learning, Asian Journal of Management Science and Applications, Vol. 3, No. 1, pp. 61 - 74, 2017, 4.
  6. (6) Regularized Distance Metric Learning for the Document Classification and its Application, J. Japan Industrial Management Association, Vol. 66, No. 2E, pp. 190 - 203, 2015, 7.
  7. (7) Multi-valued Document Classification based on coding theory, China-USA Business Review, Vol. 12, No. 9, pp. 911 - 917, 2013, 9.
  8. (8) An Optimal Weighting Method in Supervised Learning of Linguistic Model for Text Classification, Industrial Engineering & Management Systems, Vol.11 No. 1, pp.87-93, 2012, 1.
Award (1) Encouragement Paper Award of AAMSA2020, Asian Journal of Management Science and Applications
(2) Best Paper Award, Japan Industrial Management Association, 2016
Grant-in-Aid for Scientific Research Support: Japan Society for Promotion of Science (JSPS)
Recruitment of research assistant(s) No
Affiliated academic society (Membership type) (1) IEEE (Regular member)
(2) IEICE: the Institute of Electronics, Information and Communication Engineers (Regular member)
(3) IPSJ: Information Processing Society of Japan (Regular member)
(4) JIMA: Japan Industrial Management Association (Regular member)
Education Field (Undergraduate level) Information Theory, Computer Networks
Education Field (Graduate level) Information Network and Security