Prognostic of glial tumours

We deal with the problem of finding a suitable model to predict survival of patients suffering from glial tumours as a function of several covariates. Estimation is based upon a retrospective study on 192 patients. Data were collected in the Hospital of Bordeaux and are analysed by Commenges and Dartigues¹ using a Cox model. In the present paper we use dynamic Bayesian models which allow effects of the covariates to change with time through a stochastic structure. The survival function at one year is also calculated as a function of the covariates with the highest prognostic values and two factors (linear combinations of the covariates) are identified which synthesize information related to the general state of the patient (age, first symptom, etc.) and the characteristics of the tumour (diameter, localization, etc.), respectively. Survival at one year is then calculated as function of the two factors. Results are reported in tabular and graphic forms.     

Comparative regression analysis coupled with rational verification for forecasting night cooling

A set of linear regression equations has been developed for forecasting night-time minimum temperatures. The predictors used in each regression are the dry-bulb and wet-bulb temperatures recorded at various times during the day. This set of regression equations can be used to begin estimating the expected minimum temperature early in the day and then update the estimate in the course of the day. It is shown that these regression equations are quite insensitive to small deviations in the input data and some special cases are discussed. The usefulness of simple or multiple regression is also examined. Under certain conditions, linear regression on hygrometrically derived humidity parameters have been transformed into non-linear expressions of dry and wet-bulb temperatures. It is shown by an example that these non-linear relationships may fit the data equally as well as the linear relationships. Comparative verification of the scheme has shown that successive updating of the predicted minimum temperature has certain advantages. Questions associated with the representativeness of the data base are also discussed. The effect of the turbulent mixing by the wind on the night cooling is incorporated in the scheme as a further upgrading of the method.     

Multiagent bayesian forecasting of structural time-invariant dynamic systems with graphical models

Time series are found widely in engineering and science. We study forecasting of stochastic, dynamic systems based on observations from multivariate time series. We model the domain as a dynamic multiply sectioned Bayesian network (DMSBN) and populate the domain by a set of proprietary, cooperative agents. We propose an algorithm suite that allows the agents to perform one-step forecasts with distributed probabilistic inference. We show that as long as the DMSBN is structural time-invariant (possibly parametric time-variant), the forecast is exact and its time complexity is exponentially more efficient than using dynamic Bayesian networks (DBNs). In comparison with independent DBN-based agents, multiagent DMSBNs produce more accurate forecasts. The effectiveness of the framework is demonstrated through experiments on a supply chain testbed.     

Value-at-Risk via mixture distributions reconsidered

Value-at-Risk (VaR) has evolved as one of the most prominent measures of downside risk in financial markets. Zhang and Cheng [M.-H. Zhang, Q.-S. Cheng, An Approach to VaR for capital markets with Gaussian mixture, Applied Mathematics and Computation 168 (2005) 1079–1085] proposed an approach to VaR for daily returns based on Gaussian mixtures, which have become rather popular in empirical economics and finance since the seminal paper of Hamilton [J.D. Hamilton, A new approach to the economic analysis of nonstationary time series and the business cycle, Econometrica 57 (2) (1989) 357–384]. However, they do not conduct tests to assess the accuracy of the mixture-implied VaR measures. Recently, Guidolin and Timmermann [M. Guidolin, A. Timmermann, Term structure of risk under alternative econometric specifications, Journal of Econometrics, 131 (2006) 285–308] showed that Markov mixture models do well in measuring VaR at a monthly frequency, but the results may not hold for daily returns due to their more pronounced non-Gaussian features. This paper provides an extensive application of various Markov mixture models to VaR for daily returns of major European stock markets, including out-of-sample backtesting. To accommodate the properties of daily returns, we consider both Gaussian and Student’s t mixtures, and we compare the performance of both uni- and multivariate models under different parameter updating schemes. We find that a univariate mixture of two Student’s t distributions performs best overall. However, by the example of the recent turmoil in financial markets, we also highlight a weak point of the approach.     

多因素灰色预测模型及其应用

为避免传统的单个因素的灰色预测的缺点,将灰色预测与多元回归相结合,提出了基于GM(1,1)的多元回归模型.并将模型应用于天津市人才预测状况,取得了较好的预测效果.     

光谱油样分析监测技术中的神经网络预测方法

光谱油样分析是机械磨损状态监测与故障诊断的重要技术,基于光谱数据的机械状态预测有利于发现机械系统的早期磨损故障。由于神经网络对于非线性模型的辨识和非平稳信号的预测,与传统预测模型相比具有明显的优势,文章将神经网络预测方法运用于光谱分析,提出了基于神经网络预测的光谱分析监测技术。在预测模型中采用了三层BP网络模型,针对神经网络的结构对于信号预测或模型辨识的精度具有影响很大的问题,文章利用遗传算法,对神经网络输入节点数、隐层节点数和网络收敛的均方误差(MSE)目标值进行了优化,得到了最优的网络预测模型。最后,对某发动机实际的光谱分析数据进行了预测和分析,并与传统ARMA模型的预测结果进行了比较,结果充分表明了本方法的有效性和优越性。     

黄河小浪底水库抗震安全监测预测问题

根据小浪底工程的实际特点 ,本文提出了黄河小浪底水利枢纽工程施工和运行期间抗震安全监测预测的一条思路。并对人工水体出现前、震情平稳期间、出现水库诱发地震期间、近场发生强烈地震期间等各阶段的工作提出了具体设想 ,准备在小浪底水库遥测地震台网开始实施.