• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.1078-1083
• /
• 2004

We proposed a method to compensate incomplete observations and made a study of environmental problem, water quality of Tama-River in Tokyo.The method is based on interpolations of the multi-layer neural networks. We call the approach as CQSAR method .which can compensate the defect data.The water quality data include defects which will give wrong effect to other normal data. The CQSAR method suppresses the wrong effect .Thus, we believe that the proposed CQSAR method has practical usability for environment examinations.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.935-940
• /
• 2005

We are faced with water pollutions on a population explosion. Considering the importance, we research European rivers based on OECD reports. Observations in the reports have defects that make evaluation of environmental situations be difficult. By using interpolations in the compensation quantitative structure-activity relation ships (CQSAR), we complement the defects in the water quality of rivers through big cities. Thus, we get complete data set for dissolved oxygen, biochemical oxygen demand, and total phosphorus. Using the data set, we examine re-naturalization of the Rhein and the Donau in Germany. We investigate the effect of dams between Slovakia and Hungary, by using reconstructions of neural networks in CQSAR. The reconstructions have functions to extract a principal relation. On the investigation, we examine assertions of conservation groups. As the result, we confirm the re-naturalization is effective, and find a negative effect of the dam construction on changes of dissolved oxygens in the Hungary Donau. We investigate the Seine and the Thames, too.