中国股票市场技术分析非线性预测能力的实证检验

运用前向人工神经网络方法对我国股票市场技术分析非线性预测能力进行了实证检验.发现基于移动平均规则的人工神经网络模型具有明显高于AR模型和各种移动平均规则线性模型的样本外预测能力.为解释技术分析方法具有非线性预测能力的原因,本文构建了一个基于异质市场假说的移动平均规则非线性模型,发现该模型的预测能力远高于其它非线性模型.表明我国股票市场存在异质性特征,技术分析方法能捕捉到不同类型投资者之间非线性的相互作用关系可能正是其具有非线性预测能力的原因.     

机构投资者与股市泡沫的形成

机构投资者"高换手率"现象引发许多争议。本文测算证券价格泡沫与检验机构投资者的微结构交易数据后发现,机构投资者的交易导致证券价格泡沫,放大了内在价值信息的反应,导致市盈率因为机构投资者的交易而高估。进一步研究证券收益的序列相关性发现,机构投资者的交易造成的价格变动符合Campbell,Grossman和Wang(1993)^[1]模型的噪音过程,即机构投资者的交易所造成的证券价格变动和证券内在价值变动无关,具体表现在证券价格短期上涨与后续期间证券价格的反转,股价呈序列负相关性。本文的研究发现证明,机构投资者的交易风格具有"投机"特征,导致了证券价格偏离内在价值,引发股市泡沫。     

基于符号收益和跳跃变差的高频波动率模型

基于符号收益率的视角,对现有的HAR-RV类及其跳跃扩展模型进行相应分解,构建新型的HAR-RV类波动率模型.进一步,结合符号收益和不同的跳跃识别检验方法,提出了包含符号跳跃变差的HAR-RV类模型,并利用样本外滚动窗预测技术和"模型信度设定"(MCS)检验法评价了各种新旧HAR-RV模型对我国沪深300股价指数波动的预测能力.结果表明:基于C_TZ跳跃识别检验的符号跳跃变差能显著改善波动率模型的短期预测能力,但在中长期波动预测时,符号跳跃变差未能明显提升HAR-RV类模型的预测精度;新提出的HAR-S-RV-TJ-TSJV模型和HAR-S-RV-TJ模型分别在对短期(未来1天)和中长期(未来5天和20天)的波动预测检验中,展现出了最高的预测精度.     

基于SVR-ARMA组合模型的日旅游需求预测

短期微观旅游需求具有强非线性特征,单一的模型很难做出准确预测.针对此问题,本文分析了著名风景区黄山2010年旅游旺季(4-10月)相关日数据的特征,在此基础上建立SVR-ARMA组合模型,用SVR模型先对原始非线性数据预测,再对SVR模型预测所产生的线性残差用ARMA模型预测,将两部分预测值几何相加得最终的预测值.最后分莉与单一的SVR和ARMA模型对比,结果表明该组合模型有更高更稳健的预测精度,很适合短期微观旅游需求.     

双长记忆GARCH族模型的预测能力比较研究——基于沪深股市数据的实证分析

GARCH族模型在金融风险的度量中有着广泛的应用。在考虑股市收益率和波动率序列双长记忆性的基础上,基于上证综合指数和深圳成份指数的日收盘价序列,从证券投资风险量化的角度,引入受险值VaR和相对正确符号指标PCS作为模型预测误差衡量指标,比较分析了双长记忆GARCH族模型在不同分布假设情况下的的拟合与预测精度。结果显示:偏t分布能较好描述沪深股市的厚尾特征;在较小的VaR水平下ARFIMA(2,d₁,0)-FIAPARCH(1,d₂,1)-skt模型对股市波动风险具有较强的预测能力,而ARFIMA(2,d₁,0)-HYGARCH(1,d₂,1)-skt对股市的涨跌趋势具有较强的预测能力。     

基于模糊信息分配理论的短期股价涨跌模式识别

基于分形市场假说的股价并不完全反映所有信息的观点,认为历史股价信息是不完备的群体型模糊信息,提出了线性信息分配条件下的信息守恒定理,建立了基于模糊信息分配理论的短期股价涨跌预测的模糊模式识别模型,通过对上证综合指数日线数据的短期预测,表明该模型具有能够动态捕捉股价短期分布特征、有效描述股价序列内蕴的短期非线性因果关系,进而具有较高的股价涨跌识别精度,并提出了金融市场收益率可能性分布的概念.     

基于RS与ANN的上市公司财务困境预测模型的实证研究

本文以中国上市公司作为研究对象,采用粗糙集理论(RS)客观选出预测模型指标体系,以因财务状况异常而被列为特别处理公司(ST公司)作为界定上市公司的财务困境标志,采用人工神经网络(ANN)寻找最佳的利用公开财务数据预测中国上市公司财务困境的模型。我们的研究结果表明,总资产报酬率等18个包括现金流量类指标的财务指标有较强的区分财务困境公司和财务健康公司的能力;行业类型和资产规模对于上市公司财务困境预测具有至关重要的作用;运用ANN建立的神经网络模型有较强和较稳定的预测能力。     

我国股票市场的非线性研究——基于LSTAR模型

本文基于LSTAR模型对我国股票市场上证A指的非线性进行了实证研究,研究结果表明我国上证A指具有明显的非线性逻辑转换的特征,该模型很好的刻画了上证A指动态周期行为,在预测上也明显的优于传统线性模型.     

中国股票市场可预测性研究：基于机器学习的视角

股市风险溢价是金融学中的一个经典研究问题.常见的线性模型存在着模型误设和参数不稳定的问题，难以有效预测风险溢价.本研究从机器学习的视角重新检视了中国股票市场的可预测性.基于1996年1月—2019年12月的数据，构建提升回归树(boosting regression trees, BRT)模型对股市收益率与波动率进行样本外预测，并构建了最优风险资产配置模型.实证结果显示：1)提升回归树方法能够对收益率、波动率和最优风险资产权重做出准确预测；2)在收益率预测中最重要的三个变量分别是净权益增加值、换手率和股价方差；挖掘预测变量之间的非线性关系是BRT预测能力的来源；3)结合提升回归树预测构建的最优风险资产组合可以为投资者带来更高的收益和效用.本研究将机器学习方法引入股票市场风险溢价的研究，为此问题的研究提供了全新的视角.     

基于主体内禀特征的股市投资网络模型及鲁棒性研究

基于网络节点属性特征的建模方法,针对股市投资者投资能力和股票标的物价格关联的内禀特征,遵循股市投资主体间的所有权关联机制,构建了股市投资广义网络及其衍生网络模型,并对网络度分布、簇系数和平均路径长度等统计参数及网络的鲁棒性特征进行了模拟仿真分析。研究发现,股市投资广义网络和衍生网络均具有无标度特征和小世界特性,但差异性显著。股市投资衍生网络节点度呈现双幂律分布,且簇系数大小基本不受网络规模大小影响;平均路径长度随网络规模的对数增长而线性增加。此外,股市投资衍生网络在随机攻击策略下的鲁棒性较高,蓄意攻击策略下的鲁棒性较低。     

A linear formulation of the two-level DEA model

In efficiency measurement, the input and output factors of similar characteristics can be grouped into input and output categories, respectively, using a weighted-average approach under the data envelopment analysis (DEA) framework to form a two-level DEA model. The resulting two-level model is nonlinear. This note transforms the nonlinear model into a linear one using a variable substitution technique. Linear models are easier to solve than their nonlinear counterparts. The linear transformation is applicable to the two-level model built from both the primal and dual forms of the conventional one-level DEA model. The linear model transformed from the nonlinear model developed from the primal form has a dual which is equivalent to the nonlinear model developed from the dual form.     

A Search for Thresholds and Other Nonlinearities in the Relationship Between Hexavalent Chromium and Lung Cancer

The exposure-response relationship for airborne hexavalent chromium exposure and lung cancer mortality is well described by a linear relative rate model. However, categorical analyses have been interpreted to suggest the presence of a threshold. This study investigates nonlinear features of the exposure response in a cohort of 2,357 chemical workers with 122 lung cancer deaths. In Poisson regression, a simple model representing a two-step carcinogenesis process was evaluated. In a one-stage context, fractional polynomials were investigated. Cumulative exposure dose metrics were examined corresponding to cumulative exposure thresholds, exposure intensity (concentration) thresholds, dose-rate effects, and declining burden of accumulated effect on future risk. A simple two-stage model of carcinogenesis provided no improvement in fit. The best-fitting one-stage models used simple cumulative exposure with no threshold for exposure intensity and had sufficient power to rule out thresholds as large as 30 microg/m3 CrO3 (16 microg/m3 as Cr+6) (one-sided 95% confidence limit, likelihood ratio test). Slightly better-fitting models were observed with cumulative exposure thresholds of 0.03 and 0.5 mg-yr/m3 (as CrO3) with and without an exposure-race interaction term, respectively. With the best model, cumulative exposure thresholds as large as 0.4 mg-yr/m3 CrO3 were excluded (two-sided upper 95% confidence limit, likelihood ratio test). A small departure from dose-rate linearity was observed, corresponding to (intensity)0.8 but was not statistically significant. Models in which risk-inducing damage burdens declined over time, based on half-lives ranging from 0.1 to 40 years, fit less well than assuming a constant burden. A half-life of 8 years or less was excluded (one-sided 95% confidence limit). Examination of nonlinear features of the hexavalent chromium-lung cancer exposure response in a population used in a recent risk assessment supports using the traditional (lagged) cumulative exposure paradigm: no intensity (concentration) threshold, linearity in intensity, and constant increment in risk following exposure.     

Prediction and Impact of Sea Level Rise on Properties and Infrastructure of Washington,DC

The city of Washington, District of Columbia (DC) will face flooding, and eventual geographic changes, in both the short‐ and long‐term future because of sea level rise (SLR) brought on by climate change, including global warming. To fully assess the potential damage, a linear model was developed to predict SLR in Washington, DC, and its results compared to other nonlinear model results. Using geographic information systems (GIS) and graphical visualization, analytical models were created for the city and its underlying infrastructure. Values of SLR used in the assessments were 0.1 m for the year 2043 and 0.4 m for the year 2150 to model short‐term SLR; 1.0 m, 2.5 m, and 5.0 m were used for long‐term SLR. All necessary data layers were obtained from free data banks from the U.S. Geological Survey and Washington, DC government websites. Using GIS software, inventories of the possibly affected infrastructure were made at different SLR. Results of the analysis show that low SLR would lead to a minimal loss of city area. Damages to the local properties, however, are estimated at an assessment value of at least US$2 billion based on only the direct losses of properties listed in real estate databases, without accounting for infrastructure damages that include military installations, residential areas, governmental property, and cultural institutions. The projected value of lost property is in excess of US$24.6 billion at 5.0 m SLR.     

Monotonic Bayesian Semiparametric Benchmark Dose Analysis

Quantitative risk assessment proceeds by first estimating a dose‐response model and then inverting this model to estimate the dose that corresponds to some prespecified level of response. The parametric form of the dose‐response model often plays a large role in determining this dose. Consequently, the choice of the proper model is a major source of uncertainty when estimating such endpoints. While methods exist that attempt to incorporate the uncertainty by forming an estimate based upon all models considered, such methods may fail when the true model is on the edge of the space of models considered and cannot be formed from a weighted sum of constituent models. We propose a semiparametric model for dose‐response data as well as deriving a dose estimate associated with a particular response. In this model formulation, the only restriction on the model form is that it is monotonic. We use this model to estimate the dose‐response curve from a long‐term cancer bioassay, as well as compare this to methods currently used to account for model uncertainty. A small simulation study is conducted showing that the method is superior to model averaging when estimating exposure that arises from a quantal‐linear dose‐response mechanism, and is similar to these methods when investigating nonlinear dose‐response patterns.     

Carbon price forecasting with a novel hybrid ARIMA and least squares support vector machines methodology

In general, due to inherently high complexity, carbon prices simultaneously contain linear and nonlinear patterns. Although the traditional autoregressive integrated moving average (ARIMA) model has been one of the most popular linear models in time series forecasting, the ARIMA model cannot capture nonlinear patterns. The least squares support vector machine (LSSVM), a novel neural network technique, has been successfully applied in solving nonlinear regression estimation problems. Therefore, we propose a novel hybrid methodology that exploits the unique strength of the ARIMA and LSSVM models in forecasting carbon prices. Additionally, particle swarm optimization (PSO) is used to find the optimal parameters of LSSVM in order to improve the prediction accuracy. For verification and testing, two main future carbon prices under the EU ETS were used to examine the forecasting ability of the proposed hybrid methodology. The empirical results obtained demonstrate the appeal of the proposed hybrid methodology for carbon price forecasting.     

基于非线性分位数回归模型的多期VaR风险测度

多期VaR主要受到持有期及波动率两个变量的影响,并且其影响模式(线性或非线性)的确定对于准确地进行VaR风险测度至关重要。非线性分位数回归模型,能够克服线性分位数回归模型只能揭示多期VaR及其影响因素之间线性依赖关系的局限,从而提高多期VaR风险测度的准确性。结合波动模型与两个非线性分位数回归方法:QRNN和SVQR,给出了多期VaR风险测度的三类方案:波动模型法、QRNN+波动模型法、SVQR+波动模型法。选取3个股票价格指数作为研究对象,考虑了6种不同形式的波动模型,得到了18个多期VaR风险测度方法进行实证比较,结果表明:波动模型选择影响到多期VaR风险测度效果;SVQR+波动模型法略优于QRNN+波动模型法,并且两者显著优于波动模型法。     

Dummy Endogenous Variables in Weakly Separable Models

In this paper, we consider the nonparametric identification and estimation of the average effect of a dummy endogenous regressor in models where the regressors are weakly but not additively separable from the error term. The model is not required to be strictly increasing in the error term, and the class of models considered includes limited dependent variable models such as discrete choice models. Conditions are established conditions under which it is possible to identify the average effect of the dummy endogenous regressor in a weakly separable model without relying on parametric functional form or distributional assumptions and without the use of large support conditions.     

汇率货币模型的非线性协整关系检验——基于深度GRU神经网络

本文采用深度门控循环单元（GRU）神经网络探讨三种汇率货币模型（弹性价格、前瞻性和实际利率差模型）的非线性协整关系。GRU技术在深度学习中具有智能记忆、自主学习和强逼近能力等优点。为此，本文运用该技术对6组典型浮动汇率制国别数据进行了非线性Johansen协整检验。结果表明，汇率与宏观经济基本面之间存在非线性协整关系，从而说明了货币模型在非线性条件下的有效性，以及先进的深度学习工具在检验经济理论中的优势。     

考虑微观结构噪声与测量误差的波动率预测

以测量误差的分布理论为基础，本文将微观结构噪声的影响引入到测量误差的方差中，构建了包含微观结构噪声影响的HARQ-N模型。使用蒙特卡洛模拟与中国股市的高频数据对HAR、HARQ、HARQ-N模型与HAR-RV-N-CJ模型的估计和预测进行了比较，研究发现，HARQ模型和HARQ-N模型的测量误差修正项对波动率的影响系数统计显著为负，HARQ-N模型的测量误差项影响系数远大于HARQ模型，更大程度地减弱当期微观结构噪声和测量误差的影响。并且，考虑微观结构噪声和测量误差的HARQ-N模型样本内和样本外预测效果在统计上显著优于HAR模型、HARQ模型与HAR-RV-N-CJ模型。     

Reassessing Benzene Cancer Risks Using Internal Doses

Human cancer risks from benzene exposure have previously been estimated by regulatory agencies based primarily on epidemiological data, with supporting evidence provided by animal bioassay data. This paper reexamines the animal-based risk assessments for benzene using physiologically-based pharmacokinetic (PBPK) models of benzene metabolism in animals and humans. It demonstrates that internal doses (interpreted as total benzene metabolites formed) from oral gavage experiments in mice are well predicted by a PBPK model developed by Travis et al. Both the data and the model outputs can also be accurately described by the simple nonlinear regression model total metabolites = 76.4x/(80.75 + x), where x = administered dose in mg/kg/day. Thus, PBPK modeling validates the use of such nonlinear regression models, previously used by Bailer and Hoel. An important finding is that refitting the linearized multistage (LMS) model family to internal doses and observed responses changes the maximum-likelihood estimate (MLE) dose-response curve for mice from linear-quadratic to cubic, leading to low-dose risk estimates smaller than in previous risk assessments. This is consistent with the conclusion for mice from the Bailer and Hoel analysis. An innovation in this paper is estimation of internal doses for humans based on a PBPK model (and the regression model approximating it) rather than on interspecies dose conversions. Estimates of human risks at low doses are reduced by the use of internal dose estimates when the estimates are obtained from a PBPK model, in contrast to Bailer and Hoel's findings based on interspecies dose conversion. Sensitivity analyses and comparisons with epidemiological data and risk models suggest that our finding of a nonlinear MLE dose-response curve at low doses is robust to changes in assumptions and more consistent with epidemiological data than earlier risk models.