基于改进注意力机制Transformer网络的快消品销量预测方法

销量预测能为企业生产计划、仓储运输提供决策支持,使企业能更好地适应市场需求。快消品销售量受众多因素的影响,具有季节性和周期性规律,传统的线性模型难以准确的预测,本文从长时序列预测的视角,运用深度学习理论,提出了一种基于订单时序和订单频率的改进自注意力机制模型(Sequence-Frequency Transformer,SFTransformer)。首先,基于快消品订单数据构建原始数据集,采用time2vec编码处理订单时序信息,并融合订单数据的时序和频率特征在基于时序的订单数据的不同订单频率分别对应不同的注意力头来关注订单数据的订单时序特征和频率特征;使用Transformer模型架构提取特征进行长时序列预测。在真实数据集上进行对比实验,SFTransformer模型在均方误差(MSE)、平均绝对误差(MAE)、均方根误差(RMSE)3项指标上均取得了最佳性能,验证了本文所提方法的有效性。

基于夜间灯光的国家级贫困县脱贫效果监测

文章基于NPP-VIIRS夜间灯光遥感影像数据对我国国家级592个贫困县在长时间序列发展的情况展开监测研究分析。该研究借鉴可持续生计框架模型在贫困研究中的应用,选取NPP-VIIRS月度复合数据为原始数据,从长时序的角度分析了我国贫困县的时空格局演变趋势。结果表明:夜间灯光指数与多维贫困指数线性相关性最强,2012至2019年总夜间灯光亮度先下降后上升,贫困县总灯光显著增长,2019年夜间灯光值增长率是129.52%。增长率最高的县为囊谦县,增长率最低的县是尼勒克县,处于低速发展的贫困县最多,高速发展的贫困县最少,总灯光值整体向东南方向迁移。

Cloning,characterization,and expression analysis of the DEAD-box family genes,Fc-vasa and Fc-PL10a,in Chinese shrimp (Fenneropenaeus chinensis)

RNA helicases of the DEAD-box and related families are involved in various cellular processes including DNA replication, DNA repair, and RNA processing. However, the function of DEAD-box proteins in aquaculture species is poorly understood at molecular level. We obtained the full-length cDNA sequences of two genes encoding helicase-related proteins, Fc-vasa and Fc-PL10a, from the testes of Chinese shrimp, Fenneropenaeus chinensis. The two predicted amino acid sequences contain all the conserved motifs characterized by the DEAD-box family and several RGG repeats in the N-terminal regions. Homology and phylogenetic analyses indicate that they belong to the vasa and PLIO subfamilies. The three-dimensional structures of the two proteins were predicted with a homology modeling approach. Both core proteins consist of two tandem RecA-like domains similar to those of the DEAD-box RNA helicase. Using reverse transcription-polymerase chain reaction （RT-PCR） and real-time PCR we found that Fc-vasa was expressed specifically in the adult gonads. Transcription decreased in the ovary but increased in the testis during gonadal development. Fc-PL10a expression was widely distributed in the tissues we examined. Using in situ hybridization, we demonstrated that the Fc-vasa transcript is localized to the cytoplasm of the spermatogonia and oocytes. Thus, our results suggest that Fc-wasa plays an important role in germ-line development, and has utility as a germ cell lineage marker which will help to generate new insight into the origin and differentiation of germ cells as well as the regulation of reproduction in F. chinensis.

Concentration–discharge patterns of weathering products from global rivers

Quantifying the functional relationships relating river discharge and weathering products places key constraints on the negative feedback between the silicate weathering and climate. In this study we analyze the concentration–discharge relationships of weathering products from global rivers using previously compiled time-series datasets for concentrations and discharge from global rivers. To analyze the nature of the covariation between specific discharge and concentrations, we use both a power law equation and a recently developed solute production equation. The solute production equation allows us to quantify weathering efficiency, or the resistance to dilution at high runoff, via the Damkohler coefficient. These results are also compared to those derived using average concentration–discharge pairs.Both the power law exponent and the Damkohler coefficient increase and asymptote as catchments exhibit increasingly chemostatic behavior, resulting in an inverse relationship between the two parameters. We also show that using thedistribution of average concentration–discharge pairs from global rivers, rather than fitting concentration–discharge relationships for each individual river, underestimates global median weathering efficiency by up to a factor of ～10. This study demonstrates the utility of long time-series sampling of global rivers to elucidate controlling processes needed to quantify patterns in global silicate weathering rates.

Research on Seismicity and Source Parameters of Small Earthquakes in the Zipingpu Reservoir before the Wenchuan M_S 8. 0 Earthquake

Based on the digital seismic waveform data observed from regional and reservoir seismic networks,the seismicity and source parameters in the Zipingpu reservoir area from January 1,2000 to May 11,2008 before the Wenchuan MS8.0 earthquake are studied. The analysis shows a continuous activity of small earthquakes before the Wenchuan MS8. 0 earthquake. The level of seismic activity was from ML2.0 to ML3.0 in recent years. The impoundment of Zipingpu reservoir started in September 2005 and the obvious increment of seismic activities occurred in February 2008,but according to the analysis of the longer time series,the frequency and intensity of seismic activities do not exceed its range,which means it is still a fluctuation of regional seismic activities or tectonic activity. At the same time,we also studied the spatio-temporal distribution and the change of apparent stress difference in the source region before the great earthquake,and the results show that the apparent stress values of small earthquakes are generally higher than fitted values of apparent stress in the reservoir area in the nearly three months before the Wenchuan MS8.0 earthquake. It can be seen from the spatial distribution of apparent stress difference,that the epicenter of the earthquake is located at a low stress distribution area,and the relatively high apparent stress difference is at the east of the epicenter. Apparent stress increment occurred in local areas.

Is China’s Economic Growth Profitled or Wage-led?——A Theoretical Inquiry and Empirical Study on the Impact of Labor Share on Growth

The paper makes a comparative review and evaluation of different theories on impacts of the distribution of labor and capital income on economic growth as well as criticism of the mainstream theories and suggests using Bhaduri-Marglin model as the theoretical basis for relevant research. By formulating the econometric model and analyzing time series data from 1978 to 2012, the paper makes an empirical research into the impact of labor share on economic growth in China. It has discovered that the positive effect of labor share on consumption demand growth in China outweighs its negative effect on investment and net export growth, concluding that growth in China is wage-led at present. That may mean that （1） increasing labor share promotes growth and the goals of equality and efficiency supports each other,＂ （2） the demand and optimizing structure＂ lies in regulation tools. key to ＂stabilizing growth, expanding domestic establishing a set of pro-labor institution and

Symbolic representation based on trend features for knowledge discovery in long time series

The symbolic representation of time series has attracted much research interest recently. The high dimensionality typical of the data is challenging, especially as the time series becomes longer. The wide distribution of sensors collecting more and more data exacerbates the problem. Representing a time series effectively is an essential task for decision-making activities such as classification, prediction, and knowledge discovery. In this paper, we propose a new symbolic representation method for long time series based on trend features, called trend feature symbolic approximation （TFSA）. The method uses a two-step mechanism to segment long time series rapidly. Unlike some previous symbolic methods, it focuses on retaining most of the trend features and patterns of the original series. A time series is represented by trend symbols, which are also suitable for use in knowledge discovery, such as association rules mining. TFSA provides the lower bounding guarantee. Experimental results show that, compared with some previous methods, it not only has better segmentation efficiency and classification accuracy, but also is applicable for use in knowledge discovery from time series.

Dynamic multivariate multiscale entropy based analysis on brain death diagnosis

The recently introduced multivariate multiscale sample entropy(MMSE)well evaluates the long correlations in multiple channels,so that it can reveal the complexity of multivariate biological signals.The existing MMSE algorithm deals with short time series statically whereas long time series are common for real-time computation in practical use.As a solution,we novelly proposed our dynamic MMSE(DMMSE)as an extension of MMSE.This helps us gain greater insight into the complexity of each section of time series,producing multifaceted and more robust estimates than the standard MMSE.The simulation results illustrated the feasibility and well performance in the brain death diagnosis.