地学信息挖掘中RPROP和分层动量增项自适应BP算法应用研究

采用RPROP和分层动量增项自适应BP算法,从最小误差、收敛速度和运算次数方面对地球化学信息进行了研究,并进行了比较分析;针对本区域的样本数据训练结果,确定采用分层动量自适应算法进行后续预测统计工作,为矿区靶区预测提供支持.     

遗传算法结合神经网络在油气产量预测中的应用

基于遗传算法的全局搜索能力和BP算法的局部精确搜索特性,通过采用遗传算法优化神经网络的方法,将遗传算法和BP算法有机结合,做到优势互补,在提高油气产量预测精度的研究中得到了很好的应用.在对国内某中小型气田油气产量的预测中,以历史产量资料进行检验,其结果表明,提出的预测方法,预测精度明显优于BP算法,证明了这种方法的有效性和可靠性.     

基于LM算法的BP神经网络的电力负荷短期预测

通过对BP神经网络输入负荷值的归一化处理,同时采用Levenberg-Marquardt(LM)算法,建立了一个改进了的BP神经网络,同时用它来对电力系统进行短期负荷预测.LM算法有效地提高了BP神经网络的收敛速度和负荷的预测精度.仿真结果表明,改进了的BP神经网络具有很高的预测精度和较强的适用能力.     

混沌粒子群神经网络在上证指数预测中的应用

上证指数预测是一个非常复杂的非线性问题,为了提高对上证指数预测的准确性,本文采用基于混沌粒子群(CPSO)算法对BP神经网络算法改进的方法来进行预测.BP神经网络算法目前已经应用到预测、聚类、分类等许多领域,取得了不少的成果.但自身也有明显的缺点,比如易陷入局部极小值、收敛速度慢等.用混沌粒子群算法改进BP神经网络算法的基本思想是用混沌粒子群算法优化BP神经网络算法的权值和阈值,在粒子群算法中加入混沌元素,提高粒子群算法的全局搜索能力.对上证指数预测的结果表明改进后的预测方法,具有更好的准确性.     

基于二次优化BP神经网络的期货价格预测

针对BP算法存在的不足,结合神经网络、遗传算法和主成分分析的优点,提出基于二次优化BP神经网络的期货价格预测算法.初次优化采用主成分分析法对网络结构进行优化,第二次优化采用自适应遗传算法对网络参数进行优化,将经过二次优化后建立的BP神经网络模型用于期货价格预测.经仿真检验,用新方法建立的模型对期货价格进行预测,在预测的精度和速度方面都优于单纯BP神经网络模型.     

基于粒子群优化的神经网络预测模型

BP学习算法多采用梯度下降法调整权值,针对其易陷入局部极小、收敛速度慢和易引起振荡的固有缺陷,提出了一种改进粒子群神经网络算法.其基本思想是:首先采用改进粒子群优化算法反复优化BP神经网络模型的权值参数组合,再用BP算法对得到的网络参数进一步精确优化,最后用得到精确的最优参数组合进行预测.实验结果表明,该算法在股指预测中的预测性能明显提高.     

基于GRA-CS-BP算法的期货价格预测

针对在采用BP神经网络进行期货价格预测时,存在的模型结构复杂,易陷入局部极小值,模型无法收敛问题.考虑从网络结构和网络参数两个方面对BP网络模型进行优化,由此提出基于GRA-CS-BP算法的期货价格预测方法.首先用灰色关联度分析法进行输入变量筛选,找出和预测价格关联度大的重要因素作为网络输入,简化网络模型整体结构.然后采用布谷鸟算法对网络权阈值参数进行优化,将经过选择优化后建立的BP神经网络模型用于期货价格预测.仿真结果表明,新模型不仅具有更高的预测精度,同时其运行的稳定性也要好于单纯BP神经网络模型,为期货价格预测提出了一种新的方法.     

基于改进BP神经网络的电力负荷预测

在现有文献研究的基础上,对BP神经网络进行了深入研究,提出了一种新的LAFBP模型,给出了模型的标准BP算法、改进BP算法、权值和阈值的初始化方法.在此基础上,用新的LAFBP模型与传统的标准BP模型对黑龙江省巴彦县的电力负荷进行了预测.预测结果表明,新的LAFBP模型不仅克服了传统的BP模型外推效果不好的缺点,而且在模型的拟合精度、学习时间和学习次数方面明显优于传统的BP模型.     

BP-GA混合优化策略在人力资源战略规划中的应用

采用混合优化策略训练神经网络,进而实现地区人力资源数据的时间序列预测.神经网络,尤其是应用反向传播(back propagation,简称BP)算法训练的神经网络,被广泛应用于预测中.但是BP神经网络训练速度慢、容易陷入局部极值.遗传算法(genetic algorithm,简称GA)具有很好的全局寻优性.因而提出将BP和GA结合起来的混合优化策略训练神经网络,来实现人力资源数据预测.与BP算法相比,数值计算结果表明预测精度高、速度快,为地区人力资源数据的时间序列预测研究提供了一条新的途径.     

改进共轭梯度算法在矿井瓦斯含量预测中的应用

本为预测矿井瓦斯含量,根据影响矿井瓦斯含量的煤层开采深度、煤层厚度、瓦斯压力、煤的变质程度、煤层顶板岩性与煤层底板岩性等主要因素建立三层BP神经网络分析模型.针对标准BP算法存在的收敛速度慢、容易陷入局部极小等问题,从理论分析角度对共轭梯度算法和改进共轭梯度算法进行对比分析研究,并且分别用标准BP算法、共轭梯度算法和改进共轭梯度算法对BP神经网络分析模型进行训练和测试.结果表明,改进共轭梯度算法收敛速度快,预测结果相对误差保持在1%以内,并且误差波动相对平稳.因此,基于改进共轭梯度算法的BP神经网络分析模型,能够有效预测矿井瓦斯含量.     

Numerical method for estimating multivariate conditional distributions

Summary  A computational framework for estimation of multivariate conditional distributions is presented. It allows the forecast of the joint distribution of target variables in dependence on explaining variables. The concept can be applied to general distribution families such as stable or hyperbolic distributions. The estimation is based on the numerical minimization of the cross entropy, using the Multi-Level Single-Linkage global optimization method. Nonlinear dependencies of conditional parameters can be modeled with help of general functional approximators such as multi-layer perceptrons. In applications, the information about a complete distribution of forecasts can be used to quantify the reliability of the forecast or for decision support. This is illustrated on a case study concerning the spare parts demand forecast. The improvement of the forecast error due to using non-Gaussian distributions is presented in another case study concerning the truck sales forecast.     

Bay of Bengal wave forecast based on genetic algorithm: A comparison of univariate and multivariate approaches

Prediction of significant wave height (SWH) field is carried out in the Bay of Bengal (BOB) using a combination of empirical orthogonal function (EOF) analysis and genetic algorithm (GA). EOF analysis is performed on 4 years (2005–2008) of numerical wave model generated SWH field, and analyzed fields of zonal (U) and meridional (V) winds. This is to decompose the space-time distributed data into spatial modes ranked by their temporal variances. Two different variants of GA are tested. In the first one, univariate GA is applied to the time series of the first principal component (PC) of SWH in the training dataset after a filtering with singular spectrum analysis (SSA) for effecting noise reduction. The generated equations are used to carry out forecast of SWH field with various lead times. In the second method, multivariate GA is applied to the SSA filtered time series of the first PC of SWH, and time- lagged first PCs of U and V and again forecast equations are generated. Once again the forecast of SWH is carried out with same lead times. The quality of forecast is evaluated in terms of root mean square error of forecast. The results are also compared with buoy data at a location. It is concluded that the method can serve as a cost-effective alternate prediction technique in the BOB.     

A Bayesian approach to term structure modeling using heavy‐tailed distributions

In this paper, we introduce a robust extension of the three‐factor model of Diebold and Li (J. Econometrics, 130: 337–364, 2006) using the class of symmetric scale mixtures of normal distributions. Specific distributions examined include the multivariate normal, Student‐t, slash, and variance gamma distributions. In the presence of non‐normality in the data, these distributions provide an appealing robust alternative to the routine use of the normal distribution. Using a Bayesian paradigm, we developed an efficient MCMC algorithm for parameter estimation. Moreover, the mixing parameters obtained as a by‐product of the scale mixture representation can be used to identify outliers. Our results reveal that the Diebold–Li models based on the Student‐t and slash distributions provide significant improvement in in‐sample fit and out‐of‐sample forecast to the US yield data than the usual normal‐based model. Copyright © 2011 John Wiley & Sons, Ltd.     

Simultaneity and non‐linear variability in financial markets: simulation and forecasting

Non‐linear variability in financial markets can emerge from several mechanisms, including simultaneity and time‐varying coefficients. In simultaneous equation systems, the reduced‐form coefficients that determine the behaviour of jointly dependent variables are products and ratios of the original structural coefficients. If the coefficients are stochastic, the resulting multiplicative interactions will result in high degrees of non‐linearity. Processes generated in this way will scale as fractals: they will exhibit intermittent outliers and scaling symmetries, i.e. proportionality relationships between fluctuations at different separation distances. A model is specified in which both the exchange rate itself and the exchange rate residual exhibit simultaneity. The exchange rate depends on other exchange rates, while the residual depends on the other residuals. The model is then simulated using embedding noise from a t‐distribution. The simulations replicate the observed properties of exchange rates, heavy‐tailed distributions and long memory in the variance. A forecasting algorithm is specified in two stages. The first stage is a model for the actual process. In the second stage the residuals are modelled as a function of the predicted rate of change. The first and second stage models are then combined. This algorithm exploits the scaling symmetry: the residual is proportional to the predicted rate of change at separation distances corresponding to the forecast horizon. The procedure is tested empirically on three exchange rates. At a daily frequency and a 1‐day forecast horizon, two‐stage models reduce the forecast error by one fourth. At a 5‐day horizon, the improvement is 10–15 percent. At a weekly frequency, the improvement at the 1‐week horizon is on the order of 30–40 percent. Copyright © 2006 John Wiley & Sons, Ltd.     

Study on network traffic forecast model of SVR optimized by GAFSA

There are some problems, such as low precision, on existing network traffic forecast model. In accordance with these problems, this paper proposed the network traffic forecast model of support vector regression (SVR) algorithm optimized by global artificial fish swarm algorithm (GAFSA). GAFSA constitutes an improvement of artificial fish swarm algorithm, which is a swarm intelligence optimization algorithm with a significant effect of optimization. The optimum training parameters used for SVR could be calculated by optimizing chosen parameters, which would make the forecast more accurate. With the optimum training parameters searched by GAFSA algorithm, a model of network traffic forecast, which greatly solved problems of great errors in SVR improved by others intelligent algorithms, could be built with the forecast result approaching stability and the increased forecast precision. The simulation shows that, compared with other models (e.g. GA-SVR, CPSO-SVR), the forecast results of GAFSA-SVR network traffic forecast model is more stable with the precision improved to more than 89%, which plays an important role on instructing network control behavior and analyzing security situation.     

Degeneracy in infinite horizon optimization

We consider sequential decision problems over an infinite horizon. The forecast or solution horizon approach to solving such problems requires that the optimal initial decision be unique. We show that multiple optimal initial decisions can exist in general and refer to their existence as degeneracy. We then present a conceptual cost perturbation algorithm for resolving degeneracy and identifying a forecast horizon. We also present a general near-optimal forecast horizon.This material is based on work supported by the National Science Foundation under Grants ECS-8409682 and ECS-8700836.     

An Algorithm for Lead Time Demand Distributions

This paper presents a recursive algorithm to compute the lead time aggregate demand distribution from the arrival distributions of each order size. This algorithm can compute the aggregate demand distribution directly from numerical data of arrival frequencies without first fitting the data frequencies to standard distributions. When the arrival distributions do follow certain standard distributions, the aggregate demand distribution can be computed explicitly. The cases of Poisson distributions and negative binomial distributions are demonstrated as examples.     

正交各向异性功能梯度层合板稳态热传导问题的精确解

对轴对称正交各向异性功能梯度层合圆板稳态热传导问题进行精确分析.假设材料热传导率沿板厚方向按指数函数形式梯度分布,从正交各向异性功能梯度圆板稳态热传导的基本方程出发,利用分离变量法,获得了在上、下表面作用任意热分布情况下的精确解.通过数值算例的分析,指出材料性质的梯度变化、板厚边界条件等分析了对温度场分布的影响.所获得的精确结果,可以作为评价其它近似方法的标准解答.     

经验正交函数与遗传算法结合的副热带高压位势场非线性模型反演

针对副热带高压的动力预报模型难以准确构建的困难,基于T106数值预报产品500 hPa位势高度场序列,用经验正交函数（EOF）分解方法对位势场序列进行了时、空分解,引入了动力系统重构思想,以EOF分解的空间模态的时间系数序列作为动力模型变量,用遗传算法全局搜索和并行计算优势,进行了动力模型参数的优化反演,建立了客观合理的非线性动力模型．通过对动力模型积分和EOF的时、空重构,实现了副热带高压的中、长期预报．试验结果表明,本文反演的动力模型的副热带高压预报效果优于常规的数值预报产品,该研究工作为副热带高压等复杂天气系统和要素场预报提供了新的方法思路和技术途径．