软件可靠性预测的ARIMA方法研究

对基于求和自回归滑动平均模型（ARIMA模型）的软件可靠性预测方法进行了研究,提出了将软件可靠性失效数据看作时间序列,通过建立相应的ARIMA（p,d,q）模型来进行预测的方法。对该方法的基本思想、模型表述、建模流程进行了详细介绍,并依据上述方法选用Musa经典数据集中的Project SS2中的数据进行了预测,结果表明预测的准确性较高,说明该方法适用于软件可靠性预测。     

Urban Traffic Information Service Application Grid

Traffic information processing is very complicated because of dynamic, cooperative and distributed features.This paper describes the prototype system version 2.0 of Urban Traffic Information Service Application Grid (UTISAG),which is based on the previous version. In this version, a new architecture and more enhanced services are introduced.The remarkable characteristic of the new system is providing dynamic information services for travelers by grid technology.Therefore, the key research includes integrating large multi-source traffic data, forecasting route status, simulating regional traffic flow parallelly, and implementing optimum dynamic travel scheme based on massive GPS data.     

Judgmental forecasting with interactive forecasting support systems

With an increasing use of DSS/EIS, managers are often required to process information coming from a variety of sources in making a final decision. However, we have little understanding of the efficiency with which people select and use the multiple pieces of information. This issue was examined under various conditions using a DSS in a forecasting task where multiple items of information were displayed on request in an interactive manner. Results indicate that overall people underacquired information. Moreover, people often selected less-reliable information. This sub-optimal behaviour did not diminish over time (it became worse). But an aggregation DSS was helpful at the task. This suggests that people seemed to have a problem in aggregating multiple pieces of information. It was also found that the independent preparation of an initial forecast improved forecast accuracy significantly. Perhaps, forecasters may prepare the initial forecast independently and use decision aids for the subsequent tasks of the forecasting process.     

基于模糊神经网络的统一潮流控制器

为了减少先验知识对统一潮流控制器中模糊规则的设计和电力系统参数的变化对统一潮流控制器性能的影响,文中采用模糊神经网络来设计统一潮流控制器.为此首先简单介绍了统一潮流控制器的控制策略,然后阐述了自组织模糊神经网络和基于遗传算法的模糊神经网络的构造方法,接着将自组织模糊神经网络、基于遗传算法的模糊神经网络结合统一潮流控制器的控制策略应用于两种统一潮流控制器.最后通过MATLAB仿真例子来验证:这两种统一潮流控制器的设计方法的有效性.     

多变量时间序列相空间重构中参数的确定

介绍了多变量时间序列相空间重构理论.提出一种新的基于平均预测误差最小化的重构参数确定方法,阐述了该方法的算法过程及一些重要特点.此方法考虑了所有重构参数对平均预测误差的影响,能够同时确定重构系统相空间所需的恰当嵌入维数及时间延迟.最后将该方法应用于股票市场非线性动力系统的相空间重构,通过比较和分析验证了其优越性.     

基于CNN和MC的水文时间序列预测组合模型

对于水位精准的预测是预防洪涝灾害的有效措施。在深度学习不断发展的背景下,提出基于卷积神经网络和马尔科夫链的水文时间序列预测组合模型,该模型解决了现有算法未考虑站点之间空间的相关性、多维输入的时候会提高特征提取中数据重建的复杂度,以及单一模型只考虑水位时间序列线性部分而未考虑非线性部分所导致的预测精度低的问题。该组合模型首先运用卷积神经网络训练水位时间序列和降雨量时间序列对未来水位进行预测,并结合原始时间序列计算得到残差序列,再将使用马尔科夫链训练残差序列得到的残差预测结果和卷积神经网络预测的值相加得到最终的结果。实验表明,该方法与现有算法相比,在预报准确率上能够取得更好的效果。     

基于时间序列分析和卡尔曼滤波的霾预报技术

为了提高霾预报的准确率,解决时序模型的预测延时和准确率不高的问题,提出了一种基于时间序列分析和卡尔曼滤波相结合的混合霾预报算法。首先,利用图检验法和单位根检验法（ADF）检验时间序列的平稳性,通过差分运算将非平稳序列转化成平稳序列,对转化后的平稳序列进行建模;然后,将得到的模型方程作为卡尔曼滤波的状态方程和观测方程,依靠卡尔曼滤波递推性进行预报。实验结果表明,采用时间序列分析和卡尔曼滤波相结合的混合霾客观预报订正方法能有效提高霾预测精度。     

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J2EE电力营销决策支持系统的研究与实现

综合运用了J2EE、数据仓库、联机分析处理、数据挖掘等技术,建立了一个多层次分布式电力营销决策支持系统。该文阐述了该决策支持系统的体系结构,分析了从操作型数据库到数据仓库和数据集市的数据采集模式,实现了Winters模型、一元线性回归以及二元指数平滑等预测模型的设计,从而为电力营销活动提供了重要的决策支持。     

Forecasting financial time series using a methodology based on autoregressive integrated moving average and Taylor expansion

Financial time series prediction is regarded as one of the most challenging job because of its inherent complexity, and the hybrid forecasting model incorporating autoregressive integrated moving average and support vector machine (SVM) has been implemented widely to deal with the both linear and nonlinear patterns in time series data. However, the SVM model does not take into consideration the time correlation knowledge between different data points in time series, which impacts the learning efficiency of the SVM in real application. To overcome this restriction, this paper proposes the Taylor Expansion Forecasting model as an alternative to the SVM and develops a novel hybrid methodology via combining autoregressive integrated moving average and Taylor Expansion Forecasting to exploit the comprehensive forecasting capacity to the financial time series data with noise. Both theoretical proof and empirical results obtained on several commodity future prices demonstrate that the proposed hybrid model improves greatly the forecasting accuracy.