A comparison of the gravity model and the perceptron : an analytical method for perceptrons by partial differential coefficients

スポンサーリンク

概要

• 論文の詳細を見る

• Parameters of gravity models are regarded as values of partial elasticity or sensitivity of outputs to inputs. Comparable coefficients can be defined by partial differential coefficients of perceptrons if transfer functions are defined as differentiable. Elasticity was compared between a Poisson gravity model and a PR-perceptron (a customized perceptron with the Poisson error term) using the flows of inter-prefectural migration of Japan in 1990. Values of elasticity of the PR-perceptron were distributed around those of the Poisson gravity model, but exhibited wider variations. From these distributions it was recognized that travel time was the most sensitive factor for both the Poisson gravity model and the PR-perceptron, yet the PR-perceptron fits the spatial interaction better taking account of non-linear and interaction effects among input variables.

• 首都大学東京の論文

著者

• Nakaya Tomoki

  Department Of Geography Ritsumeikan University

• Nakaya Tomoki

  Department Of Geography Tokyo Metropolitan University

関連論文

• Climatic determinants of the northern range limit of Fagus crenata forests in Japan
• A comparison of the gravity model and the perceptron : an analytical method for perceptrons by partial differential coefficients
• An Analysis of Inter-Municipal Migration Flows in Japan Using GIS and Spatial Interaction Modeling

スポンサーリンク