一种基于模糊C均值聚类小数据量计算最大Lyapunov指数的新方法

在小数据量计算最大Lyapunov指数的过程中,为了减少人为因素识别线性区域带来的误差,提出一种基于模糊C均值聚类的新方法.该方法根据平均发散程度指数曲线的变化特征,利用分类算法进行识别.首先,利用小数据量算法对混沌时间序列进行计算得到平均发散程度指数集合;其次,利用模糊C均值聚类算法对平均发散程度指数集合进行分类,得到不饱和数据;然后,对不饱和的二阶差分数据进行分类,得到零附近波动数据并剔除粗大误差,再对保留的有效数据利用统计方法识别出线性区域;最后,对线性区域进行最小二乘法拟合得到最大Lyapunov指数.为了验证该算法的有效性,对著名Logistic和Hénon混沌系统进行了仿真,所得结果接近理论值.实验表明,所提出的新方法与主观识别方法比较,计算结果更加准确.

A novel method based on the fuzzy C-means clustering to calculate the maximal Lyapunov exponent from small data

In order to reduce errors caused by human factors to identify the linear region, we propose a new method based on the fuzzy C-means clustering for calculating the maximum Lyapunov exponent from small data. The method based on the changing characteristic of divergence index curve is used to identify the linear region. Firstly, the divergence index data are calculated from the small data algorithm for the given chaotic time series. Secondly, the fuzzy C-means clustering method is used for dividing the data into two classes (unsaturated and saturated data), and the unsaturated data are retained. Thirdly, the retained data are divided by the same clustering method into three classes (positive fluctuation data, zero fluctuation data and negative fluctuation data), and the zero fluctuation data are retained. Fourthly, the 3σ$ criterion is used for excluding gross errors to retain the valid from the selected data. Finally, the regression analysis and statistical test are used to identify the linear region from the valid data. The effectiveness of the proposed method can be demonstrated by the famous chaotic systems of Logistic and Henon. The calculated results are closr to the theoretical values than the subjective method. Experimental results show that the proposed new approach is easier to operate, more efficient and more accurate as compared with the subjective recognition. But this method has its own shortcomings. (1) As the new method is verified by the simulation experiment, there exists no strict mathematical proof. (2) Since the difference algorithm is used in this new method, it will miss some detailed information in some cases. (3) The calculation accuracy still needs to be improved, so this method only serves as a reference to detect the linear region, it can not be applied to high precision engineering field. Considering the deficiencies of the new method, we will make further research to improve the calculation method for maximum Lyapunovexponent, so as to make it solve the real-time problem of the signal detection, and find the accurate location of abrupt climate change in the field of meteorology, to provide accurate satellite launch safety period in the field of space weather and other aspects. In short, studying the largest Lyapunov exponent from chaotic time series has a wide application prospect and practical significance.