深圳市供水量的最优组合预测

城市供水系统是一个复杂的大系统,供水量受多种因素的共同影响。本文以深圳市最近20多年的供水量历史数据为基础,建立了深圳市供水量的最优组合预测模型。该模型具有较高的预测精度,组合预测的预测效果优于任意一种单一预测的预测效果,供水量预测结果对深圳市未来供水的短期或长期规划能起到重要的宏观指导作用。     

Analysis of call centre arrival data using singular value decomposition

We consider the general problem of analysing and modelling call centre arrival data. A method is described for analysing such data using singular value decomposition (SVD). We illustrate that the outcome from the SVD can be used for data visualization, detection of anomalies (outliers), and extraction of significant features from noisy data. The SVD can also be employed as a data reduction tool. Its application usually results in a parsimonious representation of the original data without losing much information. We describe how one can use the reduced data for some further, more formal statistical analysis. For example, a short‐term forecasting model for call volumes is developed, which is multiplicative with a time series component that depends on day of the week. We report empirical results from applying the proposed method to some real data collected at a call centre of a large‐scale U.S. financial organization. Some issues about forecasting call volumes are also discussed. Copyright © 2005 John Wiley & Sons, Ltd.     

Medium‐term horizon volatility forecasting: A comparative study

In this paper, volatility is estimated and then forecast using unobserved components‐realized volatility (UC‐RV) models as well as constant volatility and GARCH models. With the objective of forecasting medium‐term horizon volatility, various prediction methods are employed: multi‐period prediction, variable sampling intervals and scaling. The optimality of these methods is compared in terms of their forecasting performance. To this end, several UC‐RV models are presented and then calibrated using the Kalman filter. Validation is based on the standard errors on the parameter estimates and a comparison with other models employed in the literature such as constant volatility and GARCH models. Although we have volatility forecasting for the computation of Value‐at‐Risk in mind the methodology presented has wider applications. This investigation into practical volatility forecasting complements the substantial body of work on realized volatility‐based modelling in business. Copyright © 2007 John Wiley & Sons, Ltd.     

Large‐time behavior of a two‐scale semilinear reaction–diffusion system for concrete sulfatation

We study the large‐time behavior of (weak) solutions to a two‐scale reaction–diffusion system coupled with a nonlinear ordinary differential equations modeling the partly dissipative corrosion of concrete (or cement)‐based materials with sulfates. We prove that as t → ∞ , the solution to the original two‐scale system converges to the corresponding two‐scale stationary system. To obtain the main result, we make use essentially of the theory of evolution equations governed by subdifferential operators of time‐dependent convex functions developed combined with a series of two‐scale energy‐like time‐independent estimates. Copyright © 2014 John Wiley & Sons, Ltd.     

基于品牌价值的销售额预测模型应用研究

首先分析了企业的品牌价值与其销售额的变化情况,归纳出品牌价值与销售额之间的两个重要的变化规律.而后以此作为二者关系方程构建的基础假设,以广告反映模型(VW模型)为基础,将原模型中的广告投入变量替换为品牌价值变量,建立了品牌价值与销售额之间的关系方程.通过求解关系方程的平衡解,得到了以品牌价值为变量的销售额预测模型.最后,采用海尔集团的实际数据对所建模型的适用性和正确性进行了证明,同时还对模型的预测精度进行了检验.研究突破了以往研究一般采用时间、环境变量直接预测销售额,或是采用历史数据对销售额进行模拟预测等传统的预测方法,首次从品牌价值的角度出发研究其对销售额的影响途径及影响力度.品牌价值是能够包涵如广告投入等因素的全部信息量的重要指标,因此其能够更为全面客观地衡量商品在整个市场中的价值潜力、能够更准确地反映其未来的销售份额.     

FORECASTING TREE SIZE DISTRIBUTION BY NUMERICAL APPROXIMATION OF THE FORWARD KOLMOGOROV EQUATION

ABSTRACT. Forest management planning of uneven‐aged stands involves forecasting of the tree size distribution. The temporal development of the size distribution in a forest stand may be described by the forward Kolmogorov equation. The objective of this study is to illustrate that numerical approximation of the solution to the equation provides a reasonably accurate way of forecasting future tree size distribution, especially for stands with non‐normal size distribution. Furthermore, a method for the practical application is devised. The analyses compare observed and forecasted tree size distributions for two forest stands, 1) an unthinned stand of Sitka spruce (Picea sitchensis (Bong.) Carr.), and 2) an uneven‐aged stand of beech (Fagus sylvatica L.) managed under the selection system. The analyses show that the size distribution in the uneven‐aged stand may be forecasted correctly for a 20 25 year period, while for the even‐aged stand the method seems to fail after 10 to 15 years.     

MODELING AND FORECASTING MONTHLY AVERAGE WATER LEVELS BASED ON THE ELLIPTIC ORBIT MODEL

Evaluation and forecasting of water‐level fluctuation for one river is of increasing importance since it is intimately associated with human welfare and socioeconomic sustainability development. In this study, it is found that time series of monthly water‐level fluctuation exhibits annual cyclical variation. Then with annual periodic extension for monthly water‐level fluctuation, the so‐called “elliptic orbit model” is proposed for describing monthly water‐level fluctuation by mapping its time series into the polar coordinates. Experiments and result analysis indicate potentiality of the proposed method that it yields satisfying results in evaluating and forecasting monthly water‐level fluctuation at the monitoring stations in the Yangtze River of China. It is shown that the monthly water‐level fluctuation is well described by the proposed elliptic orbit model, which offers a vivid approach for modeling and forecasting monthly water‐level fluctuation in a concise and intuitive way.     

The emergence of large‐scale coherent structure under small‐scale random bombardments

We provide mathematical justification of the emergence of large‐scale coherent structure in a two‐dimensional fluid system under small‐scale random bombardments with small forcing and appropriate scaling assumptions. The analysis shows that the large‐scale structure emerging out of the small‐scale random forcing is not the one predicted by equilibrium statistical mechanics. But the error is very small, which explains earlier successful prediction of the large‐scale structure based on equilibrium statistical mechanics. © 2005 Wiley Periodicals, Inc.     

AN ADAPTIVE LEARNING FRAMEWORK FOR FORECASTING SEASONAL WATER ALLOCATIONS IN IRRIGATED CATCHMENTS

Abstract This paper describes an adaptive learning framework for forecasting end‐season water allocations using climate forecasts, historic allocation data, and results of other detailed hydrological models. The adaptive learning framework is based on artificial neural network (ANN) method, which can be trained using past data to predict future water allocations. Using this technique, it was possible to develop forecast models for end‐irrigation‐season water allocations from allocation data available from 1891 to 2005 based on the allocation level at the start of the irrigation season. The model forecasting skill was further improved by the incorporation of a set of correlating clusters of sea surface temperature (SST) and the Southern oscillation index (SOI) data. A key feature of the model is to include a risk factor for the end‐season water allocations based on the start of the season water allocation. The interactive ANN model works in a risk‐management context by providing probability of availability of water for allocation for the prediction month using historic data and/or with the incorporation of SST/SOI information from the previous months. All four developed ANN models (historic data only, SST incorporated, SOI incorporated, SST‐SOI incorporated) demonstrated ANN capability of forecasting end‐of‐season water allocation provided sufficient data on historic allocation are available. SOI incorporated ANN model was the most promising forecasting tool that showed good performance during the field testing of the model.     

基于组合预测模型的广东省第三产业产值预测

首先分析了影响广东省第三产业发展的主要因素,指出由于上述因素相互制约、相互影响,导致第三产业的发展呈现出高度的非线性特征,并使得单一的预测模型在预测效果和泛化能力方面难以胜任.在此基础上,提出了基于神经网络集成的组合预测模型,对广东省第三产业的发展进行预测,阐述了算法的基本原理和数据处理流程,实证分析表明:基于神经网络集成的组合预测模型要比单一预测模型的预测精度高.     

Bayesian tail‐risk forecasting using realized GARCH

A realized generalized autoregressive conditional heteroskedastic (GARCH) model is developed within a Bayesian framework for the purpose of forecasting value at risk and conditional value at risk. Student‐t and skewed‐t return distributions are combined with Gaussian and student‐t distributions in the measurement equation to forecast tail risk in eight international equity index markets over a 4‐year period. Three realized measures are considered within this framework. A Bayesian estimator is developed that compares favourably, in simulations, with maximum likelihood, both in estimation and forecasting. The realized GARCH models show a marked improvement compared with ordinary GARCH for both value‐at‐risk and conditional value‐at‐risk forecasting. This improvement is consistent across a variety of data and choice of distributions. Realized GARCH models incorporating a skewed student‐t distribution for returns are favoured overall, with the choice of measurement equation error distribution and realized measure being of lesser importance. Copyright © 2017 John Wiley & Sons, Ltd.     

Fixed-size Least Squares Support Vector Machines: A Large Scale Application in Electrical Load Forecasting

Based on the Nyström approximation and the primal-dual formulation of the least squares support vector machines, it becomes possible to apply a nonlinear model to a large scale regression problem. This is done by using a sparse approximation of the nonlinear mapping induced by the kernel matrix, with an active selection of support vectors based on quadratic Renyi entropy criteria. The methodology is applied to the case of load forecasting as an example of a real-life large scale problem in industry. The forecasting performance, over ten different load series, shows satisfactory results when the sparse representation is built with less than 3% of the available sample.     

Design of stochastic hitless‐prediction router by using the first exceed level theory

This paper deals with an enhanced hitless‐prediction router system that has the hitless‐restart capability with forecasting. Hitless‐restart means that the router can stay on the forwarding path and the network topology remains stable. But the major difficulty of the current hitless‐restart is that the router is always active to take the action, such as non‐stop forwarding (upgrade, maintenance and capacity expansion may be included as third party activities). Stochastic hitless‐prediction model gives the decision making factors that manage a router system more efficiently. An analogue of the first exceed level theory is applied for the restriction of the number of buffer size that is the router capacity. Analytically, tractable results are obtained by using a first exceed level process that enables us to determine the decision making factors such as recycle periods of the hitless‐prediction point to prevent a router shutdown. Copyright © 2005 John Wiley & Sons, Ltd.     

Deep learning for energy markets

Deep Learning (DL) is combined with extreme value theory (EVT) to predict peak loads observed in energy grids. Forecasting energy loads and prices is challenging due to sharp peaks and troughs that arise due to supply and demand fluctuations from intraday system constraints. We propose a deep temporal extreme value model to capture these effects, which predicts the tail behavior of load spikes. Deep long‐short‐term memory architectures with rectified linear unit activation functions capture trends and temporal dependencies, while EVT captures highly volatile load spikes above a prespecified threshold. To illustrate our methodology, we develop forecasting models for hourly price and demand from the PJM interconnection. The goal is to show that DL‐EVT outperforms traditional methods, both in‐ and out‐of‐sample, by capturing the observed nonlinearities in prices and demand spikes. Finally, we conclude with directions for future research.     

Artificial intelligence and social forecasting*

Artificial intelligence (AI) is the study of how to write programs enabling computers to do things that would require intelligence if done by people, and it could engage with social forecasting in two ways. First, it is part of the overall social‐technological context within which forecasters work. Commercial Al‐programs will affect markets and life‐styles; and advice‐giving “expert” systems will raise novel legal, social, and psychological problems. Second, AI‐programs might be used for making the social forecasts. Unlike the (essentially quantitative) computer models used for this purpose today, they could reason (and explain themselves) in verbal form. Writing an expert system requires clarification of the theories, assumptions, and “rule‐of‐thumb” inferences concerned. It would be easier to identify the inherent moral‐political bias than it is in models comprising sets of differential equations.     

一种新的非线性模糊自适应变权重组合预测模型

对于复杂工业系统非线性时间序列预测精度不高问题,引入了多种预测方法的预测相对误差、预测对象的变化趋势、灰色基本权重和自适应调节系数等概念,建立了模糊自适应变权重非线性组合预测模型。结果表明,此模糊自适应变权重非线性组合预测模型的精度较高,并且平均误差和预测平方根误差均较小。该组合预测模型为复杂非线性工业系统所需决策提供了有力支持。     

A Theoretically Consistent Version of the Nelson and Siegel Class of Yield Curve Models

A popular class of yield curve models is based on the Nelson and Siegel approach of ‘fitting’ yield curve data with simple functions of maturity. However, such models cannot be consistent across time. This article addresses that deficiency by deriving an intertemporally consistent and arbitrage‐free version of the Nelson and Siegel model. Adding this theoretical consistency expands the potential applications of the Nelson and Siegel approach to exercises involving a time‐series context, such as forecasting the yield curve and pricing interest rate derivatives. As a practical example, the intertemporal consistency of the model is exploited to derive a theoretical framework for forecasting the yield curve. The empirical application of that framework to United States data results in out‐of‐sample forecasts that outperform the random walk over the sample period of almost 50 years, for forecast horizons ranging from six months to three years.     

A Jacobi–Davidson method for two‐real‐parameter nonlinear eigenvalue problems arising from delay‐differential equations

The critical delays of a delay‐differential equation can be computed by solving a nonlinear two‐parameter eigenvalue problem. The solution of this two‐parameter problem can be translated to solving a quadratic eigenvalue problem of squared dimension. We present a structure preserving QR‐type method for solving such quadratic eigenvalue problem that only computes real‐valued critical delays; that is, complex critical delays, which have no physical meaning, are discarded. For large‐scale problems, we propose new correction equations for a Newton‐type or Jacobi–Davidson style method, which also forces real‐valued critical delays. We present three different equations: one real‐valued equation using a direct linear system solver, one complex valued equation using a direct linear system solver, and one Jacobi–Davidson style correction equation that is suitable for an iterative linear system solver. We show numerical examples for large‐scale problems arising from PDEs. Copyright © 2012 John Wiley & Sons, Ltd.     

Point and interval forecasting for carbon price based on an improved analysis-forecast system

Effective analysis and forecasting of carbon prices, which is an essential endeavor for the carbon trading market, is still considered a difficult task because of the nonlinearity and nonstationarity inherent in carbon prices. Previous studies have failed at the analysis and interval prediction of carbon prices and are limited to point forecasts. Therefore, an improved carbon price analysis and forecasting system that consists of an analysis module and a forecasting module is established in this study; more importantly, the forecasting module includes point forecasting and interval forecasting as well. Aimed at investigating the characteristics of the carbon price series, a chaotic analysis based on the maximum Lyapunov exponent is performed, the determination of appropriate distribution functions based on our newly proposed hybrid optimization algorithm is conducted, and different distribution functions are effectively designed in the analysis module. Furthermore, in the point forecasting model, the phase space reconstruction technique is applied to reconstruct the sequences decomposed by variational mode decomposition due to the chaotic characteristics of the carbon price series, and the reconstructed sequences are considered as the optimal input–output variables of the forecasting model. Then, an adaptive neuro-fuzzy inference system model is trained by the newly proposed hybrid optimization algorithm, which is developed for the first time in the domain of carbon price point forecasting. Moreover, based on the results of point forecasting and the distribution function of the carbon price series determined by the analysis module, the interval forecasting results can be obtained and implemented to provide more reliable information for decision making. Empirical results based on the carbon price data of the European Union Emissions Trading System and Shenzhen of China demonstrate that the proposed system achieves better results than other benchmark models in point forecasting as well as interval forecasting.     

Deep Learning for Real-Time Crime Forecasting and Its Ternarization

Real-time crime forecasting is important. However, accurate prediction of when and where the next crime will happen is difficult. No known physical model provides a reasonable approximation to such a complex system. Historical crime data are sparse in both space and time and the signal of interests is weak. In this work, the authors first present a proper representation of crime data. The authors then adapt the spatial temporal residual network on the well represented data to predict the distribution of crime in Los Angeles at the scale of hours in neighborhood-sized parcels. These experiments as well as comparisons with several existing approaches to prediction demonstrate the superiority of the proposed model in terms of accuracy. Finally, the authors present a ternarization technique to address the resource consumption issue for its deployment in real world. This work is an extension of our short conference proceeding paper [Wang, B., Zhang, D., Zhang, D. H., et al., Deep learning for real time Crime forecasting, 2017, arXiv: 1707.03340].