@techreport{oai:ipsj.ixsq.nii.ac.jp:00218639,
 author = {伊従, 寛哉 and 松島, 慎 and Hiroya, Iyori and Shin, Matsushima},
 issue = {9},
 month = {Jun},
 note = {医学研究や社会科学などの実質科学分野ではデータが順序尺度で得られることが少なくない.目的変数がこのような順序尺度で与えられるような問題を順序回帰とよび,回帰問題とも分類問題とも違った特徴を持つ.順序回帰問題の教師あり学習においては,未知のデータに対する予測性能の高さとともに,学習したモデルが解釈性を持つことも重要である.本稿では解釈性と予測性能の両方に優れた全変動正則化付きの加法モデルを順序回帰問題に対して拡張し,予測性能と解釈性の両方に優れている全変動正則化付き加法累積ロジットモデル (TVACLM) を提案する., In many fields such as medical research and social science, data on an ordinal scale are often obtained. Problems in which the target variable is given on the ordinal scale are called ordinal regression. Ordinal regression has different characteristics from those of regression and classification problems. In supervised learning of the ordinal regression problems, interpretability of the learned model is very important as well as its predictive performance. In this paper, we extend the generalized additive model with total variation regularization to ordinal regression problems and propose a additive cumulative logit model with total varition regularization (TVACLM) that achieves good performance in both perspectives from interpretability and prediction.},
 title = {順序回帰のための全変動正則化付き加法累積ロジットモデル},
 year = {2022}
}