二進木を用いる河川水質データの特徴抽出

概要

論文の詳細を見る
Water-quality data of rivers, for example, COD and BOD are important indices for environmental conservation. Whether we can distinguish between the sampling positions or the fiscal year, the sampling time of the water are depend on the way of characterizing the sampling positions or time of the water. We focused our attentions to one of the decision trees, the classification and the regressions trees (CART), because we should have clear splitting rules between the sampling positions or the sampling fiscal years. As a result of using CART, we found that classification between the upper, middle and lower area of Tama river could be completely done based on the conductivity and the chlorine ion. Furthermore, from the splitting rules between the water-quality data of fiscal year '97, '98 and '99, the important point concerning the purification of Tama river's water-quality, would be considerable reduction of the nitrogen-related chemical species (Total nitrogen, NH_4, NO_2, NO_3). CART should be a useful method for finding the meanings of classifying or the splitting rules of the complicated analytical data or environment data, which have recently increased in number.
社団法人日本分析化学会の論文
2003-05-05