相似信息优先的复杂装备费用测算模型

针对横截面数据的复杂装备费用预测问题,由于主要参数比较相似的装备系统其研制费用往往趋同,提出了横截面数据的“相似信息优先原理”。分析新研装备与以往装备的相似度,按照指标信息的相似度区分横截面信息的优先次序。用归一化的相似度作为加权最小二乘法的权重,通过实例建立相似信息优先的回归模型,并与其他方法的结果对比,结果表明本文的方法具有更高的预测精度,相似信息优先的多元回归模型充分反映了相似信息的规律。     

短时交通流预测FSMSVR模型

针对城市道路短时交通流的复杂非线性特点和以往的预测仅考虑典型交通条件（无交通事故等突发事件）的现状,结合交通流的特征,提出了一种有限状态机支持向量回归模型（finite state machine of support vector regression model,FSMSVR）的短时交通流预测机制. 通过线性回归算法和指数平滑算法划分交通流状态,根据各状态特点结合支持向量回归算法建立有限状态机工作机制,实现涵盖典型和非典型交通条件的短时交通流预测. 通过实验例证,对比了FSMSVR模型和传统SVR模型对城市道路6min交通流的预测,研究结果表明,该预测机制能够提高预测精确度,在非典型条件下有着较好的预测表现.     

Correcting the January optimism effect

Each month, various professional forecasters give forecasts for next year's real gross domestic product (GDP) growth and unemployment. January is a special month, when the forecast horizon moves to the following calendar year. Instead of deleting the January data when analyzing forecast updates, I propose a periodic version of a test regression for weak-form efficiency. An application of this periodic model for many forecasts across a range of countries shows that in January GDP forecast updates are positive, whereas the forecast updates for unemployment are negative. I document that this January optimism about the new calendar year is detrimental to forecast accuracy. To empirically analyze Okun's law, I also propose a periodic test regression, and its application provides more support for this law.     

基于协方差矩阵的主动学习方法及应用研究

由于工业过程采集的数据中常包含大量的无标签样本,而有标签样本数量少且人工标记成本较高,因此,提出一种基于协方差矩阵的主动学习方法。利用有标签样本建立高斯过程回归模型,并构建无标签样本之间的协方差矩阵,以协方差矩阵行列式的值作为评价指标。在挑选信息量较大的无标签样本的同时,衡量样本间的相似性,避免样本的冗余添加,最终在相同标记代价下提升模型预测精度。基于工业过程数据进行算法的应用仿真,验证了所提方法的可行性和有效性。     

The Information Content of Equity Block Trades on the Warsaw Stock Exchange: Conventional and Bootstrap Approaches

This paper focuses on the Polish stock market by analysing the information content of 95 equity block trade transactions executed on shares of companies constituting the WIG20 index. A normalized conventional approach and a bootstrap approach are used to draw inferences. These approaches make use of a multivariate regression model with two explanatory variables: a market return and a dummy variable for the event. Resampling allows construction of an empirical distribution of the normalized test statistic. The outcomes obtained from the application of a normalized conventional approach as well as a bootstrap approach are in line and confirm that equity block trade transactions carry an important signal to investors. Significant abnormal positive (negative) returns are associated with the execution of the equity block trades, the prices of which are higher (lower) than the closing prices 2 days before the execution of the equity block trade transactions. Copyright © 2015 John Wiley & Sons, Ltd.     

Treed Avalanche Forecasting: Mitigating Avalanche Danger Utilizing Bayesian Additive Regression Trees

Little Cottonwood Canyon Highway is a dead‐end, two‐lane road leading to Utah's Alta and Snowbird ski resorts. It is the only road access to these resorts and is heavily traveled during the ski season. Professional avalanche forecasters monitor this road throughout the ski season in order to make road closure decisions in the face of avalanche danger. Forecasters at the Utah Department of Transportation (UDOT) avalanche guard station at Alta have maintained an extensive daily winter database on explanatory variables relating to avalanche prediction. Whether or not an avalanche crosses the road is modeled in this paper via Bayesian additive tree methods. Utilizing daily winter data from 1995 to 2011, results show that using Bayesian tree analysis outperforms traditional statistical methods in terms of realized misclassification costs that take into consideration asymmetric losses arising from two types of error. Closing the road when an avalanche does not occur is an error harmful to resort owners, and not closing the road when one does may result in injury or death. Copyright © 2016 John Wiley & Sons, Ltd.     

Real‐Time Forecasts of Inflation: The Role of Financial Variables

We present a mixed‐frequency model for daily forecasts of euro area inflation. The model combines a monthly index of core inflation with daily data from financial markets; estimates are carried out with the MIDAS regression approach. The forecasting ability of the model in real time is compared with that of standard VARs and of daily quotes of economic derivatives on euro area inflation. We find that the inclusion of daily variables helps to reduce forecast errors with respect to models that consider only monthly variables. The mixed‐frequency model also displays superior predictive performance with respect to forecasts solely based on economic derivatives. Copyright © 2012 John Wiley & Sons, Ltd.     

Backward‐in‐Time Selection of the Order of Dynamic Regression Prediction Model

We investigate the optimal structure of dynamic regression models used in multivariate time series prediction and propose a scheme to form the lagged variable structure called Backward‐in‐Time Selection (BTS), which takes into account feedback and multicollinearity, often present in multivariate time series. We compare BTS to other known methods, also in conjunction with regularization techniques used for the estimation of model parameters, namely principal components, partial least squares and ridge regression estimation. The predictive efficiency of the different models is assessed by means of Monte Carlo simulations for different settings of feedback and multicollinearity. The results show that BTS has consistently good prediction performance, while other popular methods have varying and often inferior performance. The prediction performance of BTS was also found the best when tested on human electroencephalograms of an epileptic seizure, and for the prediction of returns of indices of world financial markets.Copyright © 2013 John Wiley & Sons, Ltd.     

基于主成分回归分析的瓦斯含量预测

为了增加多元回归模型预测的精度,将主成分分析与多元回归分析相结合提出了PCA—MRA模型,并将该模型用于实际瓦斯含量预测。结果表明,PCA—MRA模型消除了输入变量之间的相关性,减少了输入变量值个数,提高了预测精度,便于实际推广和应用,为瓦斯含量预测提供一种新的途径。     

Dynamic Latent Class Model Averaging for Online Prediction

We consider the problem of online prediction when it is uncertain what the best prediction model to use is. We develop a method called dynamic latent class model averaging, which combines a state‐space model for the parameters of each of the candidate models of the system with a Markov chain model for the best model. We propose a polychotomous regression model for the transition weights to assume that the probability of a change in time depends on the past through the values of the most recent time periods and spatial correlation among the regions. The evolution of the parameters in each submodel is defined by exponential forgetting. This structure allows the ‘correct’ model to vary over both time and regions. In contrast to existing methods, the proposed model naturally incorporates clustering and prediction analysis in a single unified framework. We develop an efficient Gibbs algorithm for computation, and we demonstrate the value of our framework on simulated experiments and on a real‐world problem: forecasting IBM's corporate revenue. Copyright © 2014 John Wiley & Sons, Ltd.