On Avoidance of the Over-fitting in the B-Spline Non-parametric Regression Model
Hirokazu Yanagihara, Megu Ohtaki
Abstract
Open-access reader
Hirokazu Yanagihara, Megu Ohtaki
Abstract
Open-access reader
複雑なトレンドを持つデータに対して平滑化をおこなう場合,B-スプラインノンパラメトリック回帰モデルを用いた手法は,単純でかつ短い計算時間で結果を得ることができるといった利点を持つ.しかしながら,このモデルは本来柔軟なモデルであるため,適切に用いない場合に,柔らか過ぎる適合,過剰な適合を起こす場合がある.本論文では,基底関数の個数と平滑化パラメータを決定するための情報量規準を改良することでこの過剰適合を回避する手法を提案する。
OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
複雑なトレンドを持つデータに対して平滑化をおこなう場合,B-スプラインノンパラメトリック回帰モデルを用いた手法は,単純でかつ短い計算時間で結果を得ることができるといった利点を持つ.しかしながら,このモデルは本来柔軟なモデルであるため,適切に用いない場合に,柔らか過ぎる適合,過剰な適合を起こす場合がある.本論文では,基底関数の個数と平滑化パラメータを決定するための情報量規準を改良することでこの過剰適合を回避する手法を提案する。
Key concepts: Mathematics, Parametric statistics, Regression analysis, Regression, Statistics, Spline (mechanical), Nonparametric regression, B-spline