基于技术分析和CBR的证券时间序列预测模型研究

发布时间：2018-03-03 10:06

本文选题：时间序列　切入点：技术分析　出处：《昆明理工大学》2013年硕士论文　论文类型：学位论文

【摘要】：复杂的时间序列往往蕴含了很多潜在的重要信息和事物规律,我们对这类重要的复杂数据对象进行详尽的分析后,便有可能揭示事物运动、变化和发展的内在规律。金融时向序列是金融资产收益序列的重要表现形式,例如股票、基金、外汇、金融衍生品等收益率的分时线、日线等,也是经济与金融领域中最重要的数据,因此对这类数据分析和预测在金融投资预测、决策和风险管理等方面具有重要意义。目前,各种市场分析技术应用于解释股票市场和预测市场的未来趋势,这些技术不仅需要一定程度在金融和经济学的专业知识,还要收集大量关于市场的数据,而且需要很多的计算,这对中小投资者都需要花费太多的精力。人类求解问题具有鲁棒性,处理问题的能力随着经验的增长而不断地增强,复用以前经验的方法是人类专家的一种基本而重要的解决问题方法,由于人工智能中的CBR推理技术同人类推理十分相似,所以本文提出一种基于技术分析和CBR的证券时间序列预测模型。案例推理(case based reasoning, CBR)主要包含案例检索、案例重用、案例修正和案例保存四个过程。本文首先结合给定个股的证券时间序列的技术形态特征,利用基于最大最小点实现了典型技术形态的模式识别,及识别形态模式的起止时间、成交量、MA、OBV、RSI等指标属性值的信息,以连续3个形态模式作为一个完整的案例,构建案例库并表示案例。然后,利用案例检索的相似匹配算法—NN算法,检索出与目标案例相似的已经存在于案例库中的源案例,进行相似度与入库阈值的比较,最终实现对证券时间序列未来走势的预测,并验证了该模型在理论和实际应用中的准确性及有效性。
[Abstract]:Complex time series often contain a lot of potentially important information and rules of things. After we analyze these important and complex data objects in detail, it is possible to reveal the movement of things. The inherent law of change and development. The financial time-series is an important manifestation of the series of returns on financial assets, such as the time-sharing and diurnal lines of returns such as stocks, funds, foreign exchange, financial derivatives, etc. It is also the most important data in the field of economy and finance. Therefore, the analysis and prediction of this kind of data is of great significance in financial investment prediction, decision making and risk management. At present, a variety of market analytical techniques are used to explain stock markets and predict future trends in markets that require not only a degree of expertise in finance and economics, but also the collection of a large amount of data on markets, And it takes a lot of computing, and it takes a lot of effort for small and medium investors. Human solutions are robust, and the ability to deal with them increases as experience grows. The method of reusing previous experience is a basic and important problem solving method for human experts, because the CBR reasoning technology in artificial intelligence is very similar to human reasoning. So this paper presents a forecasting model of securities time series based on technical analysis and CBR. Case-based reasoning (CBR) mainly includes four processes: case retrieval, case reuse, case correction and case preservation. The pattern recognition based on the maximum and the minimum points is used to realize the pattern recognition of typical technical forms, and the information of the starting and ending time of the recognition of the morphological pattern, the value of the RSI index such as MAOBV / RSI, etc., are used as a complete case, and the continuous three morphological patterns are taken as a complete case. The case base is constructed and the case is represented. Then, the similarity matching algorithm -NN algorithm is used to retrieve the source case which is similar to the target case in the case database, and the similarity is compared with the threshold value. Finally, the prediction of the future trend of securities time series is realized, and the accuracy and validity of the model in theoretical and practical applications are verified.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP18;F830.91

【参考文献】