Can NN-algorithms and macroeconomic data improve OLS industry returns forecasts?∗

Using an extensive range of macroeconomic indicators and a number of two-stage models mixing OLS and a non-parametric approach known as the nearest neighbour algorithm, the authors analyse the potential for improving forecasts of US industry returns over those built by OLS on industry-specific variables only. Basic performance is measured by the average cross-sectional correlation over time between the 55 forecasted returns and the realized returns across industries. Since investors and asset managers typically want a steady performance over time, the volatility of this cross-sectional correlation is further taken into account in an adaptation of the Sharpe Ratio. Strong evidence is found in favour of certain macroeconomic factors as dominant industry return predictors, and some two-stage models based either purely on OLS or a mix between OLS and the non-linear model can lift both cross-sectional forecasting correlation and Sharpe Ratio. However, viewed overall in relation to the benchmark OLS model, performance is not consistently improved by any particular model.     

Multivariate Cox Hidden Markov models with an application to operational risk

ABSTRACT

Modeling multivariate time-series aggregate losses is an important actuarial topic that is very challenging due to the fact that losses can be serially dependent with heterogeneous dependence structures across loss types and business lines. In this paper, we investigate a flexible class of multivariate Cox Hidden Markov Models for the joint arrival process of loss events. Some of the nice properties possessed by this class of models, such as closed-form expressions, thinning properties and model versatility are discussed in details. We provide the expectation-maximization (EM) algorithm for efficient model calibration. Applying the proposed model to an operational risk dataset, we demonstrate that the model offers sufficient flexibility to capture most characteristics of the observed loss frequencies. By modeling the log-transformed loss severities through mixture of Erlang distributions, we can model the aggregate losses. Finally, out-of-sample testing shows that the proposed model is adequate to predict short-term future operational risk losses.     

城市干道交通拥堵识别技术研究

为快速有效地进行城市干道的交通拥堵识别,文中提出一种基于朴素贝叶斯的城市干道交通拥堵识别算法。最后,基于南京市主干道的交通调查数据,对朴素贝叶斯算法以及基于径向基函数神经网络的城市干道交通拥堵识别算法进行对比。结果表明,朴素贝叶斯算法在对城市干道交通状态的识别上比基于径向基函数神经网络算法具有更好的准确性、优越性以及更低的误判率。     

基于无交叉遗传算法的物流车辆路径问题研究

设计并实现了一种求解基本车辆路径问题的无交叉遗传算法,阐述了无交叉遗传算法求解物流车辆路径问题的关键技术,包括编码方案、初始种群生成方案、适应度函数、种群选择方案和变异操作等;并对比了使用目标函数倒数法和目标函数做差法构造适应度函数时的求解效果。实验结果表明,给出的无交叉遗传算法在求解车辆路径问题上具有良好的性能。     

岩土边坡稳定性的遗传算法分析

非线性优化方法主要缺陷在于当边坡土层为非均质复杂条件时,无法保证搜索到安全系数的全局最小解,而只能搜索到局部极小解。基于遗传算法,建立了搜索岩土边坡稳定性分析最小安全系数和滑移面中心坐标与半径的数值方法。问题的可行解在变量搜索区间内搜索,包括滑弧的圆心坐标和半径。数值模拟结果表明,遗传算法搜索到的边坡稳定最小安全系数与理论解是一致的。     

APPLYING A GENETIC ALGORITHM TO INTERNATIONAL DIVERSIFICATION OF EQUITY PORTFOLIOS: A SOUTH AFRICAN INVESTOR PERSPECTIVE

This paper uses the genetic algorithm (GA) approach to generate a portfolio optimisation scenario of a South African investor who seeks to maximise return from investing in S&P500, FTSE100, NASDAQ, DOWJONES, CAC40 and the DAX from January 1, 2005, to January 31, 2008, but facing exchange rate risk. The GA searches for the optimal solution in the entire set of financial constraints without looking for partial derivatives of the utility function. Whereas most financial problems require a non‐linear and time‐varying model, the GA, with its survival principle of offspring chromosomes, is better suited to this type of problem than local optimisation methods. The performance of the GA is compared with two non‐linear models, namely the quadratic mean‐variance (QMV), which maximises the portfolio mean‐variance, and the quadratic variance minimisation (QVM), which minimises the portfolio variance. The results show that neither the QMV nor the QVM takes into account the domestic investors' risk attitude towards investing in foreign equities and therefore does not provide any international diversification benefits. In addition, the bootstrapping scenario of 10,000 simulations reveals that neither the QMV nor the QVM outperforms the GA in terms of Sharpe ratio and flexibility in dealing with investors' risk attitude towards investing in foreign equities denominated in foreign currencies.     

网络考试系统中的命题库设计与组卷算法——以“计算机应用基础”课程为例

以＂计算机应用基础＂课程的网络考试系统开发为基础,讨论了使用面向对象分析方法建立与改进命题库数据模型并将其映射为关系模型的过程。该命题库是网络考试系统的关键组成部分即题库子系统中的核心内容。最后,在网络考试系统组卷功能的实现方面,给出了随机组卷算法的具体实现。     

基于改进的BP-LM在中国人口老龄化预测中的应用

由于人口老龄化率的预测具有高度非线性特征，这与BP神经网络能够处理非线性问题的特征相符合，但BP神经网络算法易使解陷入局部极小。基于L-M算法的改进BP神经网络可以有效克服这一问题，而且收敛速度快。通过具体的仿真及实践结果验证了改进BP的有效性，并对未来五年的中国老龄化率进行了预测。     

一类多车场集送货一体化问题的最优调度研究

针对江浙沪一带中小型配送中心提供集送货一体化服务增多、运营成本增加这一现象,通过建立该配送类型的车辆调度模型,然后利用分派启发式算法和改进的节约算法设计混合启发式算法来解决这一类问题。并给出算例,验证该算法的可操作性。     

编组站车流接续优化模型及算法

在解编顺序给定的条件下,建立车流接续优化模型。在模型构建中,列车属性方面同时考虑了换重、换长和满轴三个约束,通过算例证明,该设定不但更加符合实际,而且使得车流接续优化具有了一定的灵活性。在算例分析部分,分别利用一般数学软件Lingo11.0以及VC++编写的自适应免疫克隆算法对模型进行求解,证明了模型和算法的有效性,为编组站阶段计划车流接续优化智能化提供了较好的解决途径。