多商品多策略程式交易績效曲線比較和分群機制之研究

The comparison and clustering mechanism of multi-commodity and multi-strategy program trading performance

林鈺翔(Yu-siang LIN)

指導教授 ： 許智誠

國立中央大學/管理學院/資訊管理學系/碩士(2015年)

摘要

程式交易是技術指標與程式的結合，利用兩者來控制進出場的時機。程式交易搭配上商品就會有績效的產出，這些績效是評斷一個策略好壞的基礎，市面上有一些針對程式交易產出績效的軟體，通常都是針對單一商品加上單一策略的績效，缺少績效指標的統整並呈現的方法，也沒有一個分群機制幫助了解多績效的比較和績效曲線，投資者想要了解的並不再只是簡單的賺賠而已。
由於市面上策略繁多，投資者會想要比較的並非再是單一策略加上單一商品的績效，而是多商品、多策略產出多筆績效的比較。市面上的軟體雖然可以達到多績效報表的呈現，但是大量的報表和數字類型績效的產出會讓投資者變得難以觀察。故本研究利用資訊儀表板架構建立一查詢和呈現績效的介面，讓使用者能夠方便的比較多商品多策略的績效，並且使用K-means分群演算法和小波離散轉換建立一分群機制，幫助使用者了解績效曲線的形狀，達到建立統一的查詢和動態新增比較績效的互動式介面、統一和多元的多績效比較呈現介面、績效曲線分群機制之目的，解決投資者觀察績效可能會遇到的各種問題。

關鍵字

程式交易 ； 資訊儀表板 ； 時間序列 ； 資料探勘 ； 分群機制

並列摘要

Program trading is the combination of technical index and programs, used both to control the time of entry and exit. Program trading plus commodity will have performance output, these performances are the basis to judge a strategy. There are some program trading performance output software, but it is usually for single commodity or single strategy performance output, lack of integration of performance index and performance display methods, nor a clustering mechanism to help understand performance comparison and the curves of performance. What investors want to know is no longer simply the profit of program trading.
Due to so many strategies on the market, what investors want to compare is no longer the performance of single strategy and single commodity, but the performance of multi-strategy and multi-commodity. Software now available can display multi-performance by report form, but a large number of reports and numeric type performance outputs will make investors hard to understand the performance. This study uses information dashboard architecture to build an interface of query and performance display, so that users can easily compare multi-commodity and multi-strategy performance. And uses discrete wavelet transform and k-means clustering algorithm to establish a clustering mechanism to help users understand the shape of the performance curve, and achieve the establishment of a new interactive interface of unified query and dynamic performance adding, a unity and diversity of multi-performance comparison interface and a performance curve clustering mechanism, to solve the problems investors may encounter when observing the performance.

參考文獻

(11) Mohamed Abdelfattah, 2013, A Comparison of Several Performance Dashboards Architectures, Intelligent Information Management, 35-41.

(6) 廖原豐，2006， 因果關聯規則挖掘，中央大學資訊管理學系，碩士論文。

(2) Levy, R.A., 1966, Conceptual Foundation of Technical Analysis, Financial

(3) Liao, T. W., 2005, Clustering of time series data—a survey,Pattern, recognition, 38(11), 1857-1874.

(4) Rani, S., & Sikka, G., 2012, Recent techniques of clustering of time series data: a survey, Int. J. Comput. Appl, 52(15), 1-9.