A Land-owning Pattern Choice Model with EM algorithm

• Kobayashi Risa

  The University of Tokyo

• Hato Eiji

  The University of Tokyo

Bibliographic Information

Other Title

• EMアルゴリズムを用いた土地所有形態選択問題のモデル化

Abstract

The purpose of this paper is to suggest methods of analyzing an urban history, using statistical model as land-owning pattern choice model. In constructing the model, this paper uses Cross-nested logit model which can describe the simultaneous choice probabilities, and introduces latent variables into the model assuming that there is a heterogeneity in preferences for land-owner. The model uses Expectation-Maximization (EM) algorithm known as the typical methods of machine learning for the parameter estimation, because EM algorithm is appropriate for maximizing Log-likelihood function in latent variable models. It proposes the methods to convert historical records to data for using the parameter estimation and accomplishes database construction which can analyze the land-own pattern in the modern local city. The parameter estimation results show statistically that there is heterogeneity in the choice preferences for land-owner, and that the tax and location variables impact on the land-owning pattern choice is significantly changed as time advances.

Journal

• Journal of the City Planning Institute of Japan

  Journal of the City Planning Institute of Japan 54 (3), 1245-1252, 2019-10-25

  The City Planning Institute of Japan

Keywords

• Landownership
• Cadastre
• Cross-Nested Logit Model
• EM algorithm

Details

• CRID: 1390282752358433152
• NII Article ID: 130007741987
• DOI: 10.11361/journalcpij.54.1245
• ISSN: 21850593
• ISSN: 09160647
• Web Site: https://www.jstage.jst.go.jp/article/journalcpij/54/3/54_1245/_pdf
• Text Lang: ja
• Data Source:
  • JaLC
  • Crossref
  • CiNii Articles
  • KAKEN
• Abstract License Flag: Disallowed