不同预测时距对模式预测能力影响之研究.PDF

不同預測時距對模式預測能力影響之研究 (1) (2) 陳鴻烈 蔡大偉 摘要 本研究是以德基水庫為研究對象，利用時間序列模式來預測其未來的優養化變化趨勢，協 助該區優養化問題的經營管理。研究分為兩個部分 ，第一部分是使用三種由傳統時間序列模式 演變而成的預測模式，依據不同的「預測時距」來做優養化預測分析。第二部分則是選用三種 模式評比方法來檢視各個模式最佳時距的選擇。研究顯示，以「月」作為模式預測時距時，會 得到最佳結果，此時預測值和觀測值最為接近，亦即能呈現水庫未來最真實的優養化變化趨勢。 關鍵字( ：模式預測、時距選擇、優養化) Effects of Predictive Time Period on Modeling Predictive Ability Paris Honglay Chen Professor, Department of Soil and Water Conservation, National Chung-Hsing University, Taichung, Taiwan 402, R.O.C. David D-W. Tsai Graduate Student, Department of Soil and Water Conservation, National Chung-Hsing University, Taichung, Taiwan 402, R.O.C. ABSTRACT Using three Time Series models of predicting the trophic progression and evolution of Te-Chi reservoir, this research was to determine the effects of predictive time period on modeling predictive ability. The research was divided into two parts. The first part was to predict the reservoir’s eutrophication under the three time series models with various time periods. The second part was to determine the best predictive time period for each of the three models. The results showed that the best predictive time period is “month”. The “month” predictive values were the closest to the observed values among the predictive time periods studied, i.e. the model established by the “month” period was tracking very closely the actual trophic progression of the reservoir. We believe that the results of this research can greatly enhance the eutrophication management of the reservoir and potentially other reservoi