A refined fuzzy time series model for stock market forecasting

Time series models have been used to make predictions of stock prices, academic enrollments, weather, road accident casualties, etc. In this paper we present a simple time-variant fuzzy time series forecasting method. The proposed method uses heuristic approach to define frequency-density-based partitions of the universe of discourse. We have proposed a fuzzy metric to use the frequency-density-based partitioning. The proposed fuzzy metric also uses a trend predictor to calculate the forecast. The new method is applied for forecasting TAIEX and enrollments’ forecasting of the University of Alabama. It is shown that the proposed method work with higher accuracy as compared to other fuzzy time series methods developed for forecasting TAIEX and enrollments of the University of Alabama.     

An independent test of methods of detecting system states and bifurcations in time-series data

We present an independent test of recently developed methods of potential analysis and degenerate fingerprinting which aim, respectively, to identify the number of states in a system, and to forecast bifurcations. Several samples of modelled data of unknown origin were provided by one author, and the methods were used by the two other authors to investigate these properties. The main idea of the test was to investigate whether the techniques are capable to identify the character of the data of unknown origin, which includes potentiality, possible transitions and bifurcations. Based on the results of the analysis, models were proposed that simulated data equivalent to the test samples. The results obtained were compared with the initial simulations for critical evaluation of the performance of the methods. In most cases, the methods successfully detected the number of states in a system, and the occurrence of transitions between states. The derived models were able to reproduce the test data accurately. However, noise-induced abrupt transitions between existing states cannot be forecast due to the lack of any change in the underlying potential.     

基于Lorenz模型的集合预报与单一预报的比较研究

根据非线性局部Lyapunov向量方法和增长模繁殖方法,选取Lorenz63模型和Lorenz96模型的不同状态为例,对集合预报与单一预报的预报技巧开展了对比研究.结果表明:与单一预报比较,集合预报的均方根误差和型异常相关有明显改善,随预报时间推移,改善效果越显著,且集合平均优于单一预报的实验个例数逐渐增多.就概率分布(f)而言,单一预报状态的f与真实状态基本一致,不随时间变化;而集合平均预报状态的f则随时间呈现出值域变窄、峰值变大的特点.表明随预报时间的延长,单一预报状态为混沌吸引子上的随机状态,而集合平均预报状态为吸引子子集上的随机状态,这可能是集合平均误差小于单一预报的原因.     

The application of neural networks to forecast fuzzy time series

Fuzzy time series models have been applied to handle nonlinear problems. To forecast fuzzy time series, this study applies a backpropagation neural network because of its nonlinear structures. We propose two models: a basic model using a neural network approach to forecast all of the observations, and a hybrid model consisting of a neural network approach to forecast the known patterns as well as a simple method to forecast the unknown patterns. The stock index in Taiwan for the years 1991–2003 is chosen as the forecasting target. The empirical results show that the hybrid model outperforms both the basic and a conventional fuzzy time series models.     

Comparative Analysis of Different Univariate Forecasting Methods in Modelling and Predicting the Romanian Unemployment Rate for the Period 2021–2022

Unemployment has risen as the economy has shrunk. The coronavirus crisis has affected many sectors in Romania, some companies diminishing or even ceasing their activity. Making forecasts of the unemployment rate has a fundamental impact and importance on future social policy strategies. The aim of the paper is to comparatively analyze the forecast performances of different univariate time series methods with the purpose of providing future predictions of unemployment rate. In order to do that, several forecasting models (seasonal model autoregressive integrated moving average (SARIMA), self-exciting threshold autoregressive (SETAR), Holt–Winters, ETS (error, trend, seasonal), and NNAR (neural network autoregression)) have been applied, and their forecast performances have been evaluated on both the in-sample data covering the period January 2000–December 2017 used for the model identification and estimation and the out-of-sample data covering the last three years, 2018–2020. The forecast of unemployment rate relies on the next two years, 2021–2022. Based on the in-sample forecast assessment of different methods, the forecast measures root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percent error (MAPE) suggested that the multiplicative Holt–Winters model outperforms the other models. For the out-of-sample forecasting performance of models, RMSE and MAE values revealed that the NNAR model has better forecasting performance, while according to MAPE, the SARIMA model registers higher forecast accuracy. The empirical results of the Diebold–Mariano test at one forecast horizon for out-of-sample methods revealed differences in the forecasting performance between SARIMA and NNAR, of which the best model of modeling and forecasting unemployment rate was considered to be the NNAR model.     

时序建模方法在滑油光谱分析中的应用

本文讨论了时序建模方法在机械设备滑油光谱分析中的应用。通过运用ＡＲ模型对采集的航空发动机滑油光谱数据进行时序建模和预测分析,获得了满意的效果。这一成果,为机械设备的状态监控和故障预报提供了一种实用方法。     

基于周期性截断灰色系统的电力负荷预测

电力负荷预测是电力系统调度和电力生产计划制定的重要依据。电力负荷时间序列有着明显的周期性特征。传统的电力负荷预测主要侧重于预测方法的研究,而忽略了电力负荷数据周期性特性的分析,影响了预测的准确性。针对电力负荷时间序列的周期性特征,提出了一种基于周期性截断的灰色系统模型来进行电力负荷预测。该模型利用周期性截断来反映负荷数据的周期性特征,提高了预测的精度。仿真采用EUNITE Network的公开负荷数据进行算法性能的测试,并与一些主流的电力负荷预测算法：BP神经网络、极限学习机、自回归模型以及传统的灰色系统模型做比较。仿真结果表明,周期性截断的灰色系统负荷预测的归一化均方误差和绝对平均误差是最小的。周期性截断的灰色系统为电力系统负荷预测提供了一种新的有效方法。     

基于遗传优化的西太平洋副高异常活动年份的动力预报模型反演

基于2000–2010年NECR/NECP的500 hPa位势场资料, 用EOF时空分解方法和动力模型重构思想, 通过遗传算法的全局优化搜索和并行计算途径, 开展了500 hPa位势场动力预报模型反演, 建立了刻画副高活动的非线性动力预报模型, 实现了副高活动的中长期预报. 模型预报试验表明, 该模型对副高的中长期活动, 尤其是异常活动能够较好地描述和预报, 进而为副高等复杂天气系统的预报探索了新的方法思路. 关键词： 副热带高压 500 hPa位势场 动力模型反演 遗传算法     

Forecasting confined spatiotemporal chaos with genetic algorithms

A technique to forecast spatiotemporal time series is presented. It uses a proper orthogonal or Karhunen-Loève decomposition to encode large spatiotemporal data sets in a few time series, and genetic algorithms to efficiently extract dynamical rules from the data. The method works very well for confined systems displaying spatiotemporal chaos, as exemplified here by forecasting the evolution of the one-dimensional complex Ginzburg-Landau equation in a finite domain.     

基于油液光谱分析和粒子滤波的发动机剩余寿命预测研究

油液光谱分析是机械磨损状态监测、故障诊断与故障预测的重要技术,基于光谱数据的机械状态剩余寿命预测有利于实现机械系统的最优维修决策。由于机械设备越来越复杂,其健康状态的退化过程很难用线性模型来表示,而粒子滤波(particle filter, PF)对非线性非高斯系统的处理能力,与经典Kalman滤波相比具有明显的优势,文章将PF预测方法运用于光谱分析,提出了基于PF和油液光谱分析技术的设备剩余寿命预测方法。在预测模型中实现了根据设备后验分布的估计值预测其先验分布概率,建立了基于PF的多步向前长期预测模型。最后,对某发动机实际的光谱分析数据进行了预测和分析,并与传统Kalman滤波方法的预测结果进行了比较,结果充分表明了本方法的有效性和优越性。     

Application of the nonlinear time series prediction method of genetic algorithm for forecasting surface wind of point station in the South China Sea with scatterometer observations

The present work reports the development of nonlinear time series prediction method of genetic algorithm(GA) with singular spectrum analysis(SSA) for forecasting the surface wind of a point station in the South China Sea(SCS) with scatterometer observations.Before the nonlinear technique GA is used for forecasting the time series of surface wind,the SSA is applied to reduce the noise.The surface wind speed and surface wind components from scatterometer observations at three locations in the SCS have been used to develop and test the technique.The predictions have been compared with persistence forecasts in terms of root mean square error.The predicted surface wind with GA and SSA made up to four days(longer for some point station) in advance have been found to be significantly superior to those made by persistence model.This method can serve as a cost-effective alternate prediction technique for forecasting surface wind of a point station in the SCS basin.     

A granular time series approach to long-term forecasting and trend forecasting

To overcome the “curse of dimensionality” (which plagues most predictors (predictive models) when carrying out long-term forecasts) and cope with uncertainty present in many time series, in this study, we introduce a concept of granular time series which are used to long-term forecasting and trend forecasting. A technique of fuzzy clustering is used to construct information granules on a basis of available numeric data present in the original time series. In the sequel, we develop a forecasting model which captures the essential relationships between such information granules and in this manner constructs a fundamental forecasting mechanism. It is demonstrated that the proposed model comes with a number of advantages which manifest when processing a large number of data. Experimental evidence is provided through a series of examples using which we quantify the performance of the forecasting model and provide with some comparative analysis.     

Identifying financial crises in real time

Following the thermodynamic formulation of a multifractal measure that was shown to enable the detection of large fluctuations at an early stage, here we propose a new index which permits us to distinguish events like financial crises in real time. We calculate the partition function from which we can obtain thermodynamic quantities analogous to the free energy and specific heat. The index is defined as the normalized energy variation and it can be used to study the behavior of stochastic time series, such as financial market daily data. Famous financial market crashes–Black Thursday (1929), Black Monday (1987) and the subprime crisis (2008)–are identified with clear and robust results. The method is also applied to the market fluctuations of 2011. From these results it appears as if the apparent crisis of 2011 is of a different nature to the other three. We also show that the analysis has forecasting capabilities.     

浅海混响建模的声束跟踪理论

研究建立了基于声束跟踪理论的浅海混响强度计算方法和混响时间序列仿真方法。给出了混响强度计算的简要理论推导,并进行了模型计算值与实验值的比较。建立了一种混响时间序列仿真模型,给出了其实现框架和方法。结合实验数据与文献研究结果,进行了混响序列相关特性的检验与分析。结果表明:建立的混响强度计算模型能很好地进行浅海混响强度的预报,混响序列仿真模型能仿真具有不同包络分布的混响序列,且其相关特性符合实验与文献研究结果。      

Wavelet Jensen–Shannon divergence as a tool for studying the dynamics of frequency band components in EEG epileptic seizures

We develop a quantitative method of analysis of EEG records. The method is based on the wavelet analysis of the record and on the capability of the Jensen–Shannon divergence (JSD) to identify dynamical changes in a time series. The JSD is a measure of distance between probability distributions. Therefore for its evaluation it is necessary to define a (time dependent) probability distribution along the record. We define this probability distribution from the wavelet decomposition of the associated time series. The wavelet JSD provides information about dynamical changes in the scales and can be considered a complementary methodology reported earlier [O.A. Rosso, S. Blanco, A. Rabinowicz, Signal Processing 86 (2003) 1275; O.A. Rosso, S. Blanco, J. Yordanova, V. Kolev, A. Figliola, M. Schürmann, E. Ba?ar, J. Neurosci. Methods 105 (2001) 65; O.A. Rosso, M.T. Martin, A. Figliola, K. Keller, A. Plastino, J. Neurosci. Methods 153 (2006) 163]. In the present study we have demonstrated it by analyzing EEG signal of tonic–clonic epileptic seizures applying the JSD method. The display of the JSD curves enables easy comparison of frequency band component dynamics. This would, in turn, promise easy and successful comparison of the EEG records from various scalp locations of the brain.     

基于时间序列的航天器遥测数据预测算法

在航天器遥测数据预测领域,基于时间序列的预测方法有着广阔的应用前景;时间序列有一明显的特性就是记忆性,记忆性是指时间数列中的任一观测值的表现皆受到过去观测值影响;它的基本思想是根据观测数据的特点为数据建立尽可能合理的统计模型,利用模型的统计特性解释数据的统计规律,以期达到预报的目的;提出了采用模式识别和参数估计的方法,结合航天器遥测动态数据,建立关于航天器遥测数据的时序预测模型,对航天器遥测数据趋势进行检测和预报。     

基于PCA-BP神经网络的EDXRF分析测定地质样品中铁、钛元素含量的应用研究

为实现地质样品中元素含量的准确预测,提出了基于主成分分析(PCA)的改进型BP神经网络模型。采用X荧光光谱法,对新疆西天山地质样品中Fe,Ti,V,Pb和Zn等元素进行测量,将得到的X荧光计数作为输入变量,应用该模型对未知地质样品中Fe和Ti元素进行定量预测。结果表明：主成分分析与改进型BP神经网络模型取得了较好的预测效果,预测结果与化学分析值的相对误差小于3%,为地质样品元素含量预测提供了一种新型有效的方法。     

Probability distribution function and multiscaling properties in the Korean stock market

We consider the probability distribution function (pdf) and the multiscaling properties of the index and the traded volume in the Korean stock market. We observed the power law of the pdf at the fat tail region for the return, volatility, the traded volume, and changes of the traded volume. We also investigate the multifractality in the Korean stock market. We consider the multifractality by the detrended fluctuation analysis (MFDFA). We observed the multiscaling behaviors for index, return, traded volume, and the changes of the traded volume. We apply MFDFA method for the randomly shuffled time series to observe the effects of the autocorrelations. The multifractality is strongly originated from the long time correlations of the time series.     

Generation of hierarchically correlated multivariate symbolic sequences

We introduce a method to generate multivariate series of symbols from a finite alphabet with a given hierarchical structure of similarities based on the Hamming distance. The target hierarchical structure of similarities is arbitrary, for instance the one obtained by some hierarchical clustering method applied to an empirical matrix of similarities. The method that we present here is based on a generating mechanism that does not make use of mutation rate, which is widely used in phylogenetic analysis. Here we use the proposed simulation method to investigate the relationship between the bootstrap value associated with a node of a phylogeny and the probability of finding that node in the true phylogeny. The results of this analysis are compared with those obtained in the literature according to an evolutionary model with a per-symbol constant mutation rate. We observe that the relationship between the bootstrap value of a node and the probability of the corresponding clade being correct is sensitive to both the length of data series and the length of the branch connecting the node to its closest ancestor in the phylogenetic tree, whereas such a relationship is only slightly affected by the topology of the true phylogeny and by the absolute value of similarity.     

Nonlinear Time Series Forecast Using Radial Basis Function Neural Networks

In the research of using Radial Basis Function Neural Network (RBF NN) forecasting nonlinear time series, we investigate how the different clusterings affect the process of learning and forecasting. We find that k-means clustering is very suitable. In order to increase the precision we introduce a nonlinear feedback term to escape from the local minima of energy, then we use the model to forecast the nonlinear time series which are produced by Mackey-Glass equation and stocks. By selecting the k-means clustering and the suitable feedback term, much better forecasting results are obtained.