學術產出-NSC Projects

Article View/Open

Publication Export

Google ScholarTM

政大圖書館

Citation Infomation

  • No doi shows Citation Infomation
題名 投資組合信用風險、投資組合市場風險與路徑相依選擇權評價之快速蒙地卡羅演算法研究
其他題名 Fast Monte Carlo Algorithms for Portfolio Credit Risk, Portfolio Market Risk, and Valuation of Path-Dependent Options
作者 謝明華
貢獻者 國立政治大學資訊管理學系
行政院國家科學委員會
關鍵詞 投資組合
日期 2010
上傳時間 30-Aug-2012 15:51:08 (UTC+8)
摘要 使用蒙地卡羅法來計算投資組合信用風險與投資組合市場風險的風險衡量指標是一個 高計算量的計算難題。因為這樣的模擬通常是由高維度的隨機向量所驅動, 並且模擬所 關注的事件通常為偶發事件。而使用蒙地卡羅法來評價價外路徑相依選擇權, 亦會有相 同的問題。 因此, 相關的快速演算法的設計就顯得十分重要。重點抽樣 (importance sampling) 是在偶發事件模擬中常用來加速蒙地卡羅法的技巧。而在文獻中, 也有一些成 功應用在這些問題上的重點抽樣演算法。 但這些演算法通常都要求十分複雜的測度轉 換,並不易實作。 在本計劃中, 我們將提出簡易但十分有效率的重點抽樣演算法。 我們 曾成功的設計過這樣簡易但十分有效率的重點抽樣演算法在一籃子違約交換契約的評 價上。 我們將針對由高相關性隨機向量所驅動的模擬, 開發一個一般性的演算法並將此 演算法客製化到不同的應用問題上。
Computation of the risk measures of portfolio credit risk and market risk by Monte Carlo methods is a computational intensive task, because the simulation is usually driven by a high-dimensional random vector and the set of interest is usually a rare event. Monte Carlo valuation of path-dependent options which are out of money has similar problem. Therefore, efficiency improvement algorithms for these problems are highly desired. Importance sampling is a common used technique in these setting and some successful importance sampling algorithms had been proposed to speedup these computational problems. However, these algorithms demand for complicated change of measures. In this project, we plan to proposed simple but efficient importance sampling algorithms for these problems. We have successfully implemented such algorithm for valuation of basket default swap. We will first develop a general algorithm for simulations driven by highly-correlated random vector and then specialized the algorithm to various applications.
關聯 基礎研究
學術補助
研究期間:9908~ 10007
研究經費:474仟元
資料類型 report
dc.contributor 國立政治大學資訊管理學系en_US
dc.contributor 行政院國家科學委員會en_US
dc.creator (作者) 謝明華zh_TW
dc.date (日期) 2010en_US
dc.date.accessioned 30-Aug-2012 15:51:08 (UTC+8)-
dc.date.available 30-Aug-2012 15:51:08 (UTC+8)-
dc.date.issued (上傳時間) 30-Aug-2012 15:51:08 (UTC+8)-
dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/53504-
dc.description.abstract (摘要) 使用蒙地卡羅法來計算投資組合信用風險與投資組合市場風險的風險衡量指標是一個 高計算量的計算難題。因為這樣的模擬通常是由高維度的隨機向量所驅動, 並且模擬所 關注的事件通常為偶發事件。而使用蒙地卡羅法來評價價外路徑相依選擇權, 亦會有相 同的問題。 因此, 相關的快速演算法的設計就顯得十分重要。重點抽樣 (importance sampling) 是在偶發事件模擬中常用來加速蒙地卡羅法的技巧。而在文獻中, 也有一些成 功應用在這些問題上的重點抽樣演算法。 但這些演算法通常都要求十分複雜的測度轉 換,並不易實作。 在本計劃中, 我們將提出簡易但十分有效率的重點抽樣演算法。 我們 曾成功的設計過這樣簡易但十分有效率的重點抽樣演算法在一籃子違約交換契約的評 價上。 我們將針對由高相關性隨機向量所驅動的模擬, 開發一個一般性的演算法並將此 演算法客製化到不同的應用問題上。en_US
dc.description.abstract (摘要) Computation of the risk measures of portfolio credit risk and market risk by Monte Carlo methods is a computational intensive task, because the simulation is usually driven by a high-dimensional random vector and the set of interest is usually a rare event. Monte Carlo valuation of path-dependent options which are out of money has similar problem. Therefore, efficiency improvement algorithms for these problems are highly desired. Importance sampling is a common used technique in these setting and some successful importance sampling algorithms had been proposed to speedup these computational problems. However, these algorithms demand for complicated change of measures. In this project, we plan to proposed simple but efficient importance sampling algorithms for these problems. We have successfully implemented such algorithm for valuation of basket default swap. We will first develop a general algorithm for simulations driven by highly-correlated random vector and then specialized the algorithm to various applications.en_US
dc.language.iso en_US-
dc.relation (關聯) 基礎研究en_US
dc.relation (關聯) 學術補助en_US
dc.relation (關聯) 研究期間:9908~ 10007en_US
dc.relation (關聯) 研究經費:474仟元en_US
dc.subject (關鍵詞) 投資組合en_US
dc.title (題名) 投資組合信用風險、投資組合市場風險與路徑相依選擇權評價之快速蒙地卡羅演算法研究zh_TW
dc.title.alternative (其他題名) Fast Monte Carlo Algorithms for Portfolio Credit Risk, Portfolio Market Risk, and Valuation of Path-Dependent Optionsen_US
dc.type (資料類型) reporten