本文選題：股市趨勢　切入點：規(guī)劃識別　出處：《江蘇科技大學》2012年碩士論文

【摘要】：自從股票市場誕生以來，人們就想利用股票市場中提供的各個信號去分析和預測未來的股票的走勢，由此來獲取利益。從最早非理性的“裙擺理論”的預測到運用科學理論的統(tǒng)計學和圖形學的“基本面分析”和“技術(shù)性分析”等方法，這些理論在不同的時期都取得了一定的效果。在新時期，隨著計算機技術(shù)的發(fā)展，計算機技術(shù)被廣泛應用到各個領(lǐng)域，并且取得了不錯的效果。所以人們希望將這個新的技術(shù)運用到股市分析當中，能夠幫助我們在股市當中獲取利益和規(guī)避風險。 針對于此，許多學者做出了大量的研究工作，如：將神經(jīng)網(wǎng)絡、數(shù)據(jù)挖掘等人工智能方法運用到股市預測當中，并且取得了一定的效果。但是，這些方法大多是從歷史數(shù)據(jù)中，根據(jù)歷史經(jīng)驗來推測出未來價格的變動，但是股市當中存在著大量的欺騙和博弈，往往單純的根據(jù)歷史數(shù)據(jù)很難得到合理的推論結(jié)果。 本文是將規(guī)劃識別的方法運用到股市當中，規(guī)劃識別問題是屬于心理學和人工智能的一個交叉學科,它涉及到知識表達、知識推理、非單調(diào)邏輯、情景演算、人工交互、知識挖掘等多方面知識。規(guī)劃識別中有很多方法，其中規(guī)劃識別中的基于抽象策略的規(guī)劃識別是很適合股市推測的方法，，該方法主要運用在有噪聲和不確定領(lǐng)域的識別方法。它可以建立了多層推理，在概率推理層應用馬爾科夫鏈模型作為概率推理。然后由于股市投資的每種不用的投資動作都會產(chǎn)生不同的效用，因此在第二層中引入效用函數(shù)，運用決策理論方法，來獲取較為更為合理的決策。最后可利用策略迭代層來降低計算復雜度。本文最后通過實例對策略規(guī)劃方法的有效性進行了驗證，并取得了較好的效果，通過對本文的總結(jié)和分析，對規(guī)劃識別在股市分析的前景進行展望。
[Abstract]:Since the birth of the stock market, people have wanted to use the signals provided in the stock market to analyze and predict the future trend of stocks. From the prediction of the earliest irrational "skirt theory" to the "fundamental analysis" and "technical analysis" of statistics and graphics using scientific theory, These theories have achieved certain results in different periods. In the new era, with the development of computer technology, computer technology has been widely used in various fields. So people hope that applying this new technology to stock market analysis can help us gain profits and avoid risks in the stock market. In view of this, many scholars have done a lot of research work, such as: neural network, data mining and other artificial intelligence methods used in stock market forecasting, and achieved certain results. Most of these methods are based on historical data, according to historical experience to extrapolate future price changes, but there are a lot of deception and game in the stock market, it is often difficult to get reasonable corollary results simply based on historical data. This paper applies the method of planning recognition to the stock market. The problem of planning recognition is an interdisciplinary subject of psychology and artificial intelligence. It involves knowledge expression, knowledge reasoning, non-monotone logic, scenario calculus, artificial interaction. There are many methods in planning recognition, in which abstract strategy based planning recognition is very suitable for stock market speculation. This method is mainly used in the identification of noisy and uncertain fields. The Markov chain model is used as probabilistic reasoning in the probabilistic reasoning layer. Then, because every kind of investment action of stock market investment will produce different utility, the utility function is introduced in the second layer, and the decision theory method is used. Finally, the strategy iteration layer can be used to reduce the computational complexity. Finally, the effectiveness of the policy planning method is verified by an example, and good results are obtained. Through the summary and analysis of this paper, the prospect of planning identification in stock market analysis is prospected.
【學位授予單位】：江蘇科技大學
【學位級別】：碩士
【學位授予年份】：2012
【分類號】：F224;F832.51

【參考文獻】

相關(guān)碩士學位論文 前3條

1 谷赫;時間序列的數(shù)據(jù)挖掘在證券預測分析中的應用研究[D];吉林大學;2005年

2 任紅梅;基于戰(zhàn)術(shù)規(guī)劃識別的信息對抗研究[D];東北師范大學;2007年

3 李麗;誤導動作的理論研究及其在部分可觀察規(guī)劃識別中的應用[D];東北師范大學;2007年

本文編號：1684334