查詢結果分析
來源資料
頁籤選單縮合
題 名 | 有限相空間模糊近傍差分法之混沌預測:以短期交通量預測為例=Chaotic Forecasting with Confined Space Fuzzy Neighborhoods' Difference (CSFND) Approach: Case of Short-Term Traffic Forecast |
---|---|
作 者 | 藍武王; 林豐裕; 郭怡雯; | 書刊名 | 中國土木水利工程學刊 |
卷 期 | 15:3 2003.09[民92.09] |
頁 次 | 頁589-603 |
分類號 | 557.81 |
關鍵詞 | 混沌時間序列; 有限相空間模糊近傍差分法; 形態辨識; 短期交通量預測; Chaotic time series; Confined space fuzzy neighborhoods' difference; CSFND; Pattern recognition; Short-term traffic forecast; |
語 文 | 中文(Chinese) |
中文摘要 | 本研究嘗試發展新的混沌預測方法-有限相空間模糊近傍差分 (CSFND) 法。本預測方法以Sakawa等人 (1998) 發展之模糊近傍差分 (FND) 法為基礎,加入形態辨識概念,篩選「時間相近」且「空間相近」之歷史軌道,作為模糊相等推論之依據,並藉此預測短期時間序列。實證研究採美國邁阿密高速公路低、中、高流量之每1分鐘交通量各1,780筆,分別構建CSFND法及FND法,並比較兩種方法之預測準確度,結果發現本研究發展之CSFND法比Sakawa等人之FND法有較佳之預測效果。敏感度分析顯示當相空間門檻值範圍放大時,CSFND法之預測能力會降低,最後逐漸逼近FND法。 |
英文摘要 | This paper develops a confined space fuzzy neighborhoods' difference (CSFND) approach to forecast the chaotic time-series data. This approach is based on the fuzzy neighborhoods' difference (FND) method, proposed by Sakawa, et al. (1998), which only considered the "time neighborhoods." We modify the FND method by introducing confines of "space neighborhoods" together with "pattern recognition" to perform the fuzzy reasoning. In the empirical case study, we use 1,780 one-minute flow data, respectively drawn from light, medium and heavy flow conditions of US Miami Freeway, to construct and validate both CSFND approach and FND method. The comparative results show that CSFND approach obviously outperforms on the forecasting accuracy. The sensitivity analysis also proves that as the confined space is enlarged, the forecasting capability of CSFND approach will reduce and gradually converge to FND method. |
本系統中英文摘要資訊取自各篇刊載內容。