2A3-1 生体情報を対象とした予測モデル構築のための変数分類方法の提案(OS11:医療・ヘルスケア応用(2))

概要

論文の詳細を見る
In recent years, the interest of estimating risk factor using index of human signal is growing. There are several methods for constructing model to estimate a variable based on statistics. Generally, it is necessary that obtaining rigorously ruled and tested data using the design of experiments for higher accuracy estimation model. While in practical areas, applicative accuracy is required using under looser constrains because of reason above. In this paper, we propose a method for constructing model which performs high accuracy and generality. The key point of our method is to separate variables into two types which are used for prediction and homogenization. For validation we have applied the method to estimating human's Visceral Fat Area (VFA) for diagnosis metabolic syndrome in Japan. We have a quite positive result from experiment that the correlation between estimation and reference from X-ray CT is over 0.8(p<0.05) and standard deviance of error is no matter for practical use.
一般社団法人日本機械学会の論文
2011-09-01