以雲端系統為基礎之動態物理行為分析於財務交易決策支援系統

Abstract

隨著金融海嘯的來臨到渡過，投資人面對的是財富重新分配的機會。台灣屬於淺碟型市場，因此容易受到法人或鉅額投資人的影響，也因此淪為外資或其他法人進行炒作的目標。法人在進行炒作或是資產配置時，資訊會慢慢往外散佈，大眾投資人通常是最後才能得知訊息。當投資人面對與法人之間的資訊不對稱，因為不正確的策略往往造成嚴重的損失。因此本計畫提出以物理力量為基礎進行價格及籌碼面的分析，試圖從少量的資訊中挖掘出法人所造成的異常趨勢，使投資人在面對資訊不對稱的時候，能夠及早發現異常行為，透過正確的交易策略，減少嚴重的損失，達成穩定的獲利。除了上述行為發現外，本計畫亦利用人工智慧方法學，針對動態改變的金融投資環境，進行長時間的策略配置，根據時空環境的狀況提供正確的策略給投資人降低投資人對於金融投資的門檻。本計畫將以真實交易資料進行實證，用以檢驗模型的信度與效度，並透過與傳統模型的比較評估優劣。最後，將系統佈建於雲端運算上建立一個雛形系統，以服務為導向的研究流程，將運算及儲存資源進行更有效率的配置，使投資人能夠更彈性的使用不同的決策支援服務。

When the financial crisis comes and leave, general investors get a re-allocation of wealth of opportunity. Taiwan is a Shallow-plate market; it’s likely affected by institutional investors. For this reason, institutional investors will focus and speculate in the market of Taiwan. As institutional investors speculating, the information of speculation will diffuse slowly, and the general investors are the last one who got the information. When getting asymmetric information between general investors and institutional investors, making the wrong strategy will cause the loss. Therefore, this purpose uses physical theory to analyze price and volume of Taiwan market, and extract the abnormal trend made by institutional investors from less information. When it’s asymmetric information, investors could detect the abnormal behavior early. The investors make less loss and stable profit with right strategy. In addition, this purpose use artificial intelligence methodology to adaptive the model for dynamic environment. According the environment of market the model provides the right strategy to investors, and reduces investors’ burden when planning. It wills empirical evidence in the Taiwan market, and measure the model of reliability and validity. Finally, the system built a prototype in the cloud computing, and split the research process as a service. Through cloud computing that utilization of computing and storage resources will be more efficient. The investors can be more flexible with different decision-making support services.