基于符号时间序列方法的金融收益分析与预测

Analysis and Forecasting of Financial Returns Based on Symbolic Time Series Method

引入符号时间序列分析方法从大尺度的角度分析收益变化的特征,提出了确定收益变化的主要模式并预测收益水平的方法。首先将收益序列转化为符号序列,由符号序列中不同的字代表不同的收益变化模式,根据符号序列直方图,可以确定收益变化的主要模式。然后,根据各收益变化模式的概率分布,在前几个时点收益水平确定的情况下,可以推知下一个或几个时点处于不同收益水平的概率,从而实现对收益水平的预测。对上证综指、深证成指以及上证工业股指数、上证商业股指数、上证地产股指数、上证公用事业股指数共六个股票指数的收益序列进行了实证分析,确定了各指数收益的主要变化模式,并基于主要变化模式进行了收益水平的预测,从而说明了该方法的有效性和可行性。

Symbolic time series analysis method is introduced into the analysis of the characteristics of return change from the angle of large scale.The method of determining the principal return change patterns and forecasting return levels is proposed.Firstly,the return series is transformed into symbolic series.Different words in symbolic series represent different return change pattern.The principal return change patterns can be determined according to symbolic series histogram.Then,on condition that the return levels of the former several time points are determined,according to the probability distribution of various return change patterns,the probability of different return levels of the next one or several time points can be deduced to realize the forecasting of return levels.The return series of six indexes which are Shanghai composite stock,Shenzhen component stock,Shanghai industrial stock,Shanghai commercial stock,Shanghai property stock and Shanghai utility stock are analyzed.The principal return change patterns of each index are determined and the return levels based on the principal return change patterns are forecasted to prove the effectiveness and feasibility of the method.