基于ATLG混合模型的股票价格预测

股市是金融市场的重要组成部分,对股票价格预测有着重要的意义.同时,深度学习具有强大的数据处理能力,可以解决金融时间序列的复杂性所带来的问题.对此,本文提出一种结合自注意力机制的混合神经网络模型(ATLG).该模型由长短期记忆网络(LSTM)、门控递归单元(GRU)、自注意力机制构建而成,用于对股票价格的预测.实验结果表明:(1)与LSTM、GRU、RNN-LSTM、RNN-GRU等模型相比, ATLG模型的准确率更高;(2)引入自注意力机制使模型更能聚焦于重要时间点的股票特征信息;(3)通过对比,双层神经网络起到的效果更为明显.(4)通过MACD (moving average convergence and divergence)指标进行回测检验,获得了53%的收益,高于同期沪深300的收益.结果证明了该模型在股票价格预测中的有效性和实用性.     

基于TTUR的C-DCGAN机械故障诊断模型稳定训练方法

针对条件深度卷积生成式对抗网络（conditional deep convolutional generative adversarial network ,简称C-DCGAN）在训练过程中出现的不稳定性问题,提出具有随机梯度下降的双时间尺度更新规则（two time-scale update rule,简称TTUR）用于C-DCGAN机械故障诊断模型训练中,在判别器和生成器具有单独学习速率的情况下提高模型的稳定性。首先,给出了TTUR在C?DCGAN模型中收敛性证明;其次,在西储大学轴承数据集（Case Western Reserve University,简称CWUR）和实验室行星齿轮箱数据集上验证其有效性;最后,引入Jensen-Shannon 散度（Jensen-Shannon divergence,简称JSD）指标评估模型捕获到的真实数据和生成数据之间的相似度。实验结果表明,TTUR提高了C-DCGAN的学习能力,优于传统的C-DCGAN。     

考虑风电不确定性的交直流配电网低碳分布鲁棒优化调度

为增加配电网风电的消纳能力,减少碳排放,建立了一种交直流配电网低碳分布鲁棒优化调度模型。分析风电预测误差和预测出力历史数据之间的正相关性,采用混合Copula函数,建立它们之间的联合概率分布,得到风电预测误差的条件概率分布。将交直流配电网解耦为交流和直流子网,以各自综合运行成本最小为优化目标,在交流子网优化模型中引入碳交易机制,建立交直流配电网分散协调优化模型。以得到的风电预测误差的条件概率分布为参考,构建了基于K-L散度的分布鲁棒模糊集。利用拉格朗日对偶理论,将优化模型转化为单层优化目标模型,并利用交替方向乘子法进行分散协调优化求解。基于修改后33节点交直流配电网模型的仿真结果表明所提模型能有效减少配电网侧碳排放量,显著提高风电消纳能力。     

考虑心理行为的区间二元语义动态群决策方法

针对当前属性值为区间语言信息的群决策方法的不足,提出了一种考虑决策者心理行为的区间二元语义动态多属性群决策方法。定义了一种更接近决策者思维习惯的区间二元语义新的距离公式;引入时间度概念,结合考虑专家评价的综合一致性,建立优化模型求解时序权重;并基于前景理论,构建前景偏差值最小化的权重优化模型;采用基于新距离测度的交互式多属性决策（TODIM）方法计算方案的综合总体优势度,以此对方案进行排序。并以供应链协同合作伙伴的选择与评价为例,验证了所提方法的有效性和合理性。     

谚语:语言的精品——浅议英语谚语

谚语是人民群众智慧的结晶,反映了一个民族的风俗文化和道德准则.本文从英语谚语语法的来源、其语言特点及思想内容几方面对英语谚语进行了简要的评介.     

Species differentiation of chinese mollitrichosiphum (aphididae: greenideinae) driven by geographical isolation and host plant acquirement

The impact of both the uplift of the Qinghai-Tibetan Plateau (QTP) and the separation of the Taiwan and Hainan Islands on the evolution of the fauna and flora in adjacent regions has been a topic of considerable interest. Mollitrichosiphum is a polyphagous insect group with a wide range of host plants (14 families) and distributions restricted to Southeast Asia. Based on the mitochondrial Cytochrome C Oxidase Subunit I (COI) and Cytochrome b (Cytb) genes, the nuclear elongation factor-1α (EF-1α) gene, and the detailed distribution and host plant data, we investigated the species differentiation modes of the Chinese Mollitrichosiphum species. Phylogenetic analyses supported the monophyly of Mollitrichosiphum. The divergence time of Mollitrichosiphum tenuicorpus (c. 11.0 mya (million years ago)), Mollitrichosiphum nandii and Mollitrichosiphum montanum (c. 10.6 mya) was within the time frame of the uplift of the QTP. Additionally, basal species mainly fed on Fagaceae, while species that fed on multiple plants diverged considerably later. Ancestral state reconstruction suggests that Fagaceae may be the first acquired host, and the acquisition of new hosts and the expansion of host range may have promoted species differentiation within this genus. Overall, it can be concluded that geographical isolation and the expansion of the host plant range may be the main factors driving species differentiation of Mollitrichosiphum.