發(fā)布時(shí)間：2018-06-29 03:13

  本文選題：兩水平自回歸 + 非參數(shù)AC算法��； 參考：《電子科技大學(xué)》2012年碩士論文

【摘要】：長久以來，對(duì)股票的投資策略構(gòu)建的重點(diǎn)主要集中于兩個(gè)方面，一是簡單的利用移動(dòng)平均規(guī)則進(jìn)行交易，其最終目的往往不是為了進(jìn)行投資，而是為了檢驗(yàn)市場的有效性，二是構(gòu)建投資組合以求投入到實(shí)際應(yīng)用當(dāng)中。而關(guān)于股票的投資策略最近的研究則轉(zhuǎn)向了以對(duì)股市進(jìn)行預(yù)測為基礎(chǔ)的策略構(gòu)建。 本文以均線交易規(guī)則為基礎(chǔ)，構(gòu)建了一套具有較強(qiáng)操作性的投資策略，該策略的關(guān)鍵就是通過預(yù)測模型得到均線的預(yù)測值進(jìn)而得到股價(jià)的預(yù)測值，然后通過策略設(shè)定的買賣條件進(jìn)行交易。對(duì)于短期投資者來說，提前預(yù)測到均線的走勢進(jìn)而判斷股價(jià)的走勢和拐點(diǎn)，意義就尤為重大。 本文首先使用了兩水平自回歸模型進(jìn)行預(yù)測，首先使用均線數(shù)據(jù)建立上水平模型，再同時(shí)使用均線和股價(jià)數(shù)據(jù)建立下水平模型，利用上水平模型得到的均線預(yù)測值對(duì)下水平模型的預(yù)測結(jié)果進(jìn)行調(diào)控，使其在延長預(yù)測區(qū)間的同時(shí)盡量減少預(yù)測的偏差。 接著，本文使用了非參數(shù)的AC預(yù)測模型。AC算法是一種圖形擬合的非參數(shù)預(yù)測方法，有利于尋找股價(jià)的拐點(diǎn)。通過在歷史時(shí)期尋找與當(dāng)前時(shí)期狀態(tài)相近的數(shù)據(jù)，對(duì)歷史數(shù)據(jù)的延拓進(jìn)行變換組合，得到當(dāng)前的預(yù)測值。 由于股價(jià)的變動(dòng)比較劇烈，圖形缺乏規(guī)律性，故單純使用AC算法時(shí)得到的結(jié)果還不能令人滿意，于是我們將EMD引入，通過將原始數(shù)據(jù)構(gòu)成的曲線分解為若干有一定規(guī)律性的曲線，極大的提高了AC算法的預(yù)測精度和準(zhǔn)確度。 在三個(gè)預(yù)測模型的預(yù)測結(jié)果基礎(chǔ)上，，我們使用構(gòu)建的股票投資策略，任選30只股票進(jìn)行實(shí)證。結(jié)果表明，在三種均線預(yù)測模型中，基于EMD分解的AC模型具有最高的預(yù)測準(zhǔn)確性，在相同的交易策略下其虛擬操作的結(jié)果也是最好的和唯一具備現(xiàn)實(shí)投資價(jià)值的。同時(shí)通過對(duì)三種模型的實(shí)證分析，小盤股的收益率及其波動(dòng)均大于同等情況下的大盤股。因此，基于EMD分解的AC預(yù)測模型的這個(gè)交易策略具備良好的現(xiàn)實(shí)可操作性和盈利性，可作為廣大股票投資者的重要參考。
[Abstract]:For a long time, the focus of investment strategy construction on stock is mainly focused on two aspects. First, it is simple to use mobile average rules to trade. The ultimate goal is not to invest, but to test the effectiveness of the market. The two is to build an investment portfolio in order to put into practical application. Recent research on strategy has shifted to a strategy based on stock market prediction.
On the basis of the average trading rules, this paper constructs a set of investment strategy with strong operability. The key of this strategy is to get the forecast value of the average line through the prediction model and then get the forecast value of the stock price, and then deal with the trading conditions set by the strategy. For short term investors, the trend of the average line is predicted in advance. Judging the trend and inflection point of stock price is of great significance.
First, the two level autoregressive model is used to predict. First, the level model is built using the average line data, and then the horizontal model is established by using the average line and the stock price data. The prediction results of the lower level model are adjusted by the average forecast value obtained by the upper level model, so that it can be reduced as much as possible while prolonging the prediction interval. Less predicted deviations.
Then, this paper uses the non parametric AC prediction model.AC algorithm to be a non parametric prediction method of graphic fitting, which is helpful to find the turning point of the stock price. By searching for the data similar to the current state in the historical period, the continuation of the historical data is transformed and combined to get the current prediction value.
Because the change of the stock price is more violent and the graphics are not regular, the results obtained by using the AC algorithm are not satisfactory. So we introduce the EMD into the curve which is made up of the original data into some regular curves, which greatly improves the accuracy and accuracy of the AC algorithm.
On the basis of the prediction results of the three prediction models, we use the proposed stock investment strategy and choose 30 stocks to carry out an empirical study. The results show that in the three average forecasting models, the AC model based on EMD decomposition has the highest prediction accuracy. Under the same transaction strategy, the result of its virtual operation is also the best and the only one. Real investment value. At the same time, through the empirical analysis of the three models, the yield and volatility of the small stocks are greater than those in the same situation. Therefore, the AC prediction model based on EMD decomposition has good practical maneuverability and profitability, which can be used as an important reference for the stock investors.
【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2012
【分類號(hào)】：F830.91;F224

【參考文獻(xiàn)】

相關(guān)期刊論文 前9條

1 陳果,尚松浩,雷志棟;類比合成方法在干旱區(qū)內(nèi)陸河徑流量預(yù)報(bào)中的應(yīng)用[J];干旱區(qū)地理;2004年03期

2 王志剛;曾勇;李平;;中國股票市場技術(shù)分析非線性預(yù)測能力的實(shí)證檢驗(yàn)[J];管理工程學(xué)報(bào);2009年01期

3 何躍;楊劍;徐玖平;;基于GMDH的組合預(yù)測模型應(yīng)用研究[J];計(jì)算機(jī)應(yīng)用;2007年02期

4 孫碧波;移動(dòng)平均線有用嗎?——基于上證指數(shù)的實(shí)證研究[J];數(shù)量經(jīng)濟(jì)技術(shù)經(jīng)濟(jì)研究;2005年02期

5 張賓,賀昌政,余海;基于FRI的改進(jìn)相似合成算法及成都GDP預(yù)測研究[J];西南民族大學(xué)學(xué)報(bào)(人文社科版);2004年11期

6 賀昌政,張賓,俞海;自組織數(shù)據(jù)挖掘與人工神經(jīng)網(wǎng)絡(luò)方法比較研究[J];系統(tǒng)工程理論與實(shí)踐;2002年11期

7 J.-A.穆勒,劉光中;類比合成算法及對(duì)金融預(yù)測的應(yīng)用(英文)[J];運(yùn)籌學(xué)學(xué)報(bào);2002年03期

8 余峰;田益祥;李成剛;漆泉;;基于自由現(xiàn)金流量的證券投資策略及實(shí)證[J];預(yù)測;2011年02期

9 秦宇;;應(yīng)用經(jīng)驗(yàn)?zāi)B(tài)分解的上海股票市場價(jià)格趨勢分解及周期性分析[J];中國管理科學(xué);2008年S1期

相關(guān)博士學(xué)位論文 前1條

1 王強(qiáng);有理插值樣條方法及其在數(shù)字圖像處理中的應(yīng)用研究[D];合肥工業(yè)大學(xué);2007年

相關(guān)碩士學(xué)位論文 前1條

1 汪勇;開放中的中國資本市場的投資策略[D];復(fù)旦大學(xué);2010年

本文編號(hào)：2080590