标题: 运用市场轮廓理论与分群分类 分析于台指期市场之行为发现
Discovering behavior in Taiwan Index Futures Market by Using Market Profile, Classification and Clustering Analysis
作者: 彭焕淇
Peng,Huan-Chi
陈安斌
Chen,An-Pin
资讯管理研究所
关键字: 台湾指数期货;分群分析;资料探勘;市场轮廓理论;分类分析;Taiwan Index Futures;Classification analysis;Clustering analysis;Data mining;Market profile
公开日期: 2013
摘要: 本研究以一分钟K线图为基础,搭配市场轮廓理论以及成交量轮廓图
,将一天当中每分钟所发生的交易里面隐含的讯息抽出,得到整个市场在不同的时间、不同的价格中所隐含的情绪、行为,并进一步地利用分类方法(Decision Tree、Supervised Self Organized Map) 或者分群方法(K-means、Supervised Self Organized Map)将前处理过后资料进行归纳,经过分群或分类后的资料依照各自被分配至的群组给予情绪假设的标签,将标签汇出至一文字串流,再利用字频统计去计算各连续情绪假设标签所表现出来的行为,测试一市场在连续的情绪冲击下会展现出来的行为模式,最后再将这些行为模式统计后转换成交易策略进行实证。
本研究之研究方法内涵时间序列概念,发现出来之连续行为和过往之研究或者技术指标所描绘出来的一瞬间之能量有所区别,发现出市场除了会受到大幅度波动之影响外对于连续出现只小幅度物理能量冲击也会有所反应。
实验结果证明了市场并非如随机漫步(random walk)理论之说,而是市场具有逻辑并会受到心理因素的影响,并且会经由时间的酝酿将这些情绪反应在价格上。
This research based on one minute time frame candlestick, market profile theory and volume profile. we can get the mood and behavior that generate by market by Extracting the hiding information in every minutes in day trading. Then using the classification(Decision Tree、Supervised Self Organized Map) or clustering analysis(K-means、Supervised Self Organized Map) algorithm to give a label to every one minute time frame candlestick.
After we give every candlestick a label and sort them by time, use the stream analysis algorithm to find the relationship between candlesticks then we can get a invest strategy and use the data between 2010 and 2013 to backtest.
This research including the time series concept, showing that not only the big, once psychological factors change can the continuous, little psychological factors change can affect the price in the market.
Experimental results show that the market is not random walk, the market has a logical and would be affected by psychological factors, the investors mood will be brewing through time, in the end reactions in the price.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT070153428
http://hdl.handle.net/11536/74573
显示于类别:Thesis