基于已实现半协方差的投资组合优化

在传统的风险度量方法中,常见的协方差估计量并未区分资产收益的下侧风险和上侧收益,而一般的下偏矩估计量则存在非对称性和难以加总的缺点.本文引入已实现半协方差矩阵(RSCOV)作为风险度量进行波动率预测和投资组合研究.本文将RSCOV应用于两种常见的风险分散投资策略—风险平价(ERC)策略和全局方差最小(GMV)策略,并将机器学习中的在线加权集成(OWE)算法用于提升已实现波动率预测方法HAR-RV的样本外预测表现.通过研究发现,相比起已有的其他风险衡量方式,仅包含负向波动信息的下半RSCOV能够更好地被用于平衡组内各资产的风险贡献.基于A股市场2011-2018年的高频数据,本文通过实证研究发现,OWE-HARRV在月度预测步长下的效果优于HAR-RV,而下半RSCOV则能够使ERC策略以及GMV策略在保证一定平均收益的同时,降低了组合收益的极端损失.     

An ensemble of LSTM neural networks for high‐frequency stock market classification

We propose an ensemble of long–short‐term memory (LSTM) neural networks for intraday stock predictions, using a large variety of technical analysis indicators as network inputs. The proposed ensemble operates in an online way, weighting the individual models proportionally to their recent performance, which allows us to deal with possible nonstationarities in an innovative way. The performance of the models is measured by area under the curve of the receiver operating characteristic. We evaluate the predictive power of our model on several US large‐cap stocks and benchmark it against lasso and ridge logistic classifiers. The proposed model is found to perform better than the benchmark models or equally weighted ensembles.     

基于集成学习的复杂网络链路预测及其形成机制分析

为了预测节点与网络中其他现有节点之间的新连接或缺失连接，链路(边)预测近年来引发了越来越多的研究兴趣。最近已经提出各种具有不同特点的算法，以解决链路预测的问题，其中每种算法只考虑一种网络信息，从而产生片面的结果。提出基于集成学习的方法，将所有单一算法集成组合，综合考虑网络的各种信息来解决这一问题。在8个真实网络上进行了实验，利用局部拓扑索引、全局拓扑索引和推荐算法提取了17个不同的特征。结果表明，集成学习的关键性能指标——受试者工作特征曲线 (receiver operating characteristic curve， ROC)下面积(area under curve， AUC)比最佳单一算法提高2%至17%，最高达到0.9624。此外，根据度分布和随机森林得到的特征选择，分析了不同类型网络的结构与形成机制。在形成机制、网络类型和功能之间，获得了一些重要的见解:由某些确定的机制或假设导出的特征，确实是连接2个节点的内在驱动力，也正因为如此，这些特征可以用于链路预测。     

基于互补型集成经验模态分解和遗传最小二乘支持向量机的交通流量预测模型

交通流是智能交通系统中的关键组成部分,也是交通规划的重要依据。为了提高道路交通流量预测的精确性,提出一种基于互补型集成经验模态分解(complete ensemble empirical mode decomposition,CEEMD)后,采用遗传算法(genetic algorithm,GA)优化参数的最小二乘支持向量机(least square support vector machine,LSSVM)的交通流量预测模型。该模型使用互补型集成经验模态分解原始数据,将分解后的本征模态函数(intrinsic mode function,IMF)分量分别用遗传算法优化参数后的最小二乘支持向量机进行预测,叠加全部IMF分量值作为模型最终的预测结果。通过对美国加利福利亚州某高速公路一个月的交通流量数据进行训练预测,结果表明,该模型平均相对误差仅为6.51%,相较于其他模型拥有更好的预测效果,可为交通流的预测提供一定的参考。     

Multi-model ensembles in climate science: Mathematical structures and expert judgements

Projections of future climate change cannot rely on a single model. It has become common to rely on multiple simulations generated by Multi-Model Ensembles (MMEs), especially to quantify the uncertainty about what would constitute an adequate model structure. But, as Parker points out (2018), one of the remaining philosophically interesting questions is: “How can ensemble studies be designed so that they probe uncertainty in desired ways?” This paper offers two interpretations of what General Circulation Models (GCMs) are and how MMEs made of GCMs should be designed. In the first interpretation, models are combinations of modules and parameterisations; an MME is obtained by “plugging and playing” with interchangeable modules and parameterisations. In the second interpretation, models are aggregations of expert judgements that result from a history of epistemic decisions made by scientists about the choice of representations; an MME is a sampling of expert judgements from modelling teams. We argue that, while the two interpretations involve distinct domains from philosophy of science and social epistemology, they both could be used in a complementary manner in order to explore ways of designing better MMEs.     

基于改进EEMD的卷积神经网络滚动轴承故障诊断

集合经验模态分解(EEMD,ensemble empirical mode decomposition)对信号进行分解,得到的模态函数(IMF,Intrinsic model function)在2端点存在严重的发散现象,如果将分解结果直接应用到故障诊断系统中,会导致诊断的准确率下降。首先将支持向量机(SVM,support vector machine)和EEMD算法结合进行信号分解,并利用仿真信号进行可靠性分析;其次对SVM(support rector machine)-EEMD分解的分量进行选择后再分解并构建能量向量,最后和卷积神经网络结合,构建滚动轴承故障诊断模型并通过实验验证。结果表明,改进EEMD算法可以有效缓解端点发散问题,构建的故障诊断模型提高了故障诊断精度。     

具有选择与补偿机制的加权集合序贯极端学习机及其应用

针对集成在线序贯极端学习机(EOS-ELM)预测精度不高和动态适应性差的问题,提出一种具有选择与补偿机制的加权集合序贯极端学习机.该加权集合序贯极端学习机在序贯学习过程中,通过对当前预测模型精度的判断决定是否进行递推更新操作,同时为提高预测模型的动态跟踪能力,在加入新样本的同时对旧样本进行剔除;然后,利用EMD对残差序列处理后进行预测,并将初始预测结果与残差预测结果相加得到最终预测模型.通过对上证指数的预测,结果表明所提方法具有更好的泛化性能,预测精度相比EOS-ELM提高了近36.1%.     

基于SAMME+ResNet的多相码信号识别方法

针对传统多相码信号识别方法在低信噪比情况下分类精度不高、类识别率不均衡和识别方法不具有通用性的特点,提出了一种利用集成学习中的多类指数损失函数逐步添加模型(stagewise additive modeling using a multi-class exponential loss function, SAMME)算法和残差神经网络(residual neural network, ResNet)的多相码信号识别方法。通过仿真实验对5类多相码信号进行了分类识别,验证了模型的有效性,分析了不同数量基学习器对模型的影响,最后与传统分类方法进行了对比。仿真结果表明,在信噪比低于6 dB的情况下,所提方法相对于单个残差网络提高了约10%的分类精度,同时缩小了类之间识别率的差距,相对于常用的分类方法也有很大的优势。     

取用水监测点的水量计算与变化趋势分析

加强取用水监测是实施最严格水资源管理制度的重要举措.为了更有效地利用大量的取用水在线监测数据,需要对获取的监测数据进行预处理.本文首先提出应用中位数法与曲线拟合相结合的方法对取用水监测数据进行异常值检测,再用曲线拟合方法对异常值进行校正;其次,根据校正后得到的数据进行两个方面的分析,一方面是计算监测点的年取水量,另一方面是应用集成经验模态分解方法分析监测点的日取水量变化趋势;最后,以M市的16个自来水厂2016年取用水在线监测数据为例进行实证分析,结果表明,本文提出的中位数法与曲线拟合相结合的方法能够有效地检测异常值,进而再用曲线拟合方法能够更好地对异常值校正.根据校正后得到的数据进行分析发现81%的监测点年取水量相对2011年水利普查数据有所增加,个别监测点超出许可取水量较多,75%的监测点从春季到冬季日取水量变化为先增后减的抛物线趋势.     

An Adaptive Multiscale Ensemble Learning Paradigm for Nonstationary and Nonlinear Energy Price Time Series Forecasting

For forecasting nonstationary and nonlinear energy prices time series, a novel adaptive multiscale ensemble learning paradigm incorporating ensemble empirical mode decomposition (EEMD), particle swarm optimization (PSO) and least square support vector machines (LSSVM) with kernel function prototype is developed. Firstly, the extrema symmetry expansion EEMD, which can effectively restrain the mode mixing and end effects, is used to decompose the energy price into simple modes. Secondly, by using the fine‐to‐coarse reconstruction algorithm, the high‐frequency, low‐frequency and trend components are identified. Furthermore, autoregressive integrated moving average is applicable to predicting the high‐frequency components. LSSVM is suitable for forecasting the low‐frequency and trend components. At the same time, a universal kernel function prototype is introduced for making up the drawbacks of single kernel function, which can adaptively select the optimal kernel function type and model parameters according to the specific data using the PSO algorithm. Finally, the prediction results of all the components are aggregated into the forecasting values of energy price time series. The empirical results show that, compared with the popular prediction methods, the proposed method can significantly improve the prediction accuracy of energy prices, with high accuracy both in the level and directional predictions. Copyright © 2016 John Wiley & Sons, Ltd.