基于长短记忆神经网络预测的大类资产因子配置研究

大类资产配置策略为资金相对庞大的机构投资者提供了一个有效获取稳健收益的手段,通过因子收益分布及相关性的预测能更好地进行大类资产配置.将大类资产因子配置的思想与机器学习算法预测有机结合,首先筛选宏观因子及风格因子,利用长短记忆神经网络(LSTM)方法预测组合收益,得到最优因子组合;然后结合最优因子组合中蕴含的信息,修正对资产预期收益率的估计并提出了大类资产的权重配置方案.通过在全球18种大类资产上进行的算例分析表明,采用本文模型得到的配置策略相较于其它模型有更高的收益风险比、较低的年化波动以及较小的最大回撤.研究结论可为机构投资者的大类资产配置提供理论借鉴.     

基于CVaR的多目标投资组合模型

金融市场的发展与完善,以及人民收入水平的提高,使越来越多人关注金融投资并成为热点.理性的投资者总是期望风险尽可能低同时收益又尽可能高,而且希望投资的资产易于管理和管理成本低.考虑投资者多个目标的要求,将运用CVaR风险度量方法,提出一个均值—CVaR—资产数目的多目标投资组合模型,并利用多目标粒子群算法对模型进行实证分析,验证新模型的可行性和有效性,为热衷投资的投资者进行投资组合提供一个新方法.     

网络小额贷款业务个人信用风险评估——基于DNN-SMOTEENN-ExtraTrees组合模型

针对网络小额贷款业务,构建组合模型DNN-SMOTEENN-ExtraTrees评估网络小贷信用风险.首先利用SMOTEENN算法处理样本数据中“好”和“坏”样本分布极端不平衡情况,再利用极端随机数算法ExtraTrees对特征重要性进行评估并剔除无关变量,最后采用深度神经网络DNN评估网络小贷个人信用风险.通过召回率、精确度、F1值和AUC值等模型性能评价指标,与BP神经网络模型、Logistic回归及支持向量机比较,发现组合模型分类能力更显著,泛化能力更加优异,更适合数据规模大、维度高的网络小贷市场评估信用风险.     

上市公司信用风险分析模型中的变量选择

当前上市公司信用风险数据所呈现出的高维度以及高相关性的特点严重影响了信用风险模型的准确性。为此本文结合已有算法以及信用风险模型的特点设计了一种新的基于非参数的变量选择方法。通过该方法对上市公司用风险相关变量进行分析筛选可以消除数据集中包含的噪声变量以及线性相关变量。本文同时还针对该方法设计了高变量维度下最优解求解算法。文章以Logistic模型为例对上市公司信用风险做了实证分析,研究结果表明与以往的变量选择方法相比该方法可以有效的降低数据维度,消除变量间的相关性,并同时提高模型的可靠性和预测精度。     

均值—动态VaR多阶段投资组合优化研究

将动态风险度量方法运用到多阶段投资组合中,提出了具有交易成本和交易量限制的均值—动态VaR多阶段投资组合模型,并运用自创算法——离散近似迭代法求解.方法的基本思路为:首先,将模型中的连续型状态变量离散化,并将上述模型转化多阶段赋权有向图,然后,运用极大代数求出起点至终点的最长路程,即获得模型的一个可行解;最后,以该可行解为基础,继续迭代直到前后两个可行解非常接近.证明了该方法的收敛性,并以一个具体的算例,验证了该算法可以较快地计算出不同终期财富所对应的最优投资策略.     

基于Copula函数的组合资产条件相依性模型研究

条件概率分布常用来研究马尔科夫序列相依模型的构建,组合资产的相依结构受多方面的影响,资产之间的相互影响与时间上的记忆效应是组合资产两类主要的相依关系.结合条件概率的理论建立基于Copula函数相依关系模型,研究组合资产之间同期相依关系及时间上的短期相依关系,提出了模型参数的三阶段极大似然估计方法.     

基于BP算法的信用风险评价模型研究

本文利用神经网络技术建立基于 BP算法的信用风险评价模型 ,为我国某商业银行 12 0家贷款企业进行信用风险评价 ,按照企业的信用等级分为“信用好”、“信用中等”和“信用差”三个小组 .仿真结果表明 ,本文所建立的神经网络信用风险评价模型的分类准确率高于传统的参数统计分类方法——线性判别分析法的分类准确率 .文中还详细给出神经网络信用风险评价模型的网络构建方法及基于 BP网络的学习算法和步骤 .     

协整模型的配对交易策略优化

对基于协整理论的配对交易策略进行了改进.改进后的模型利用计算机能够快速循环运算的特点,循环查找最优配对组合与建仓阈值,使模型能够快速运用到各类资产及多种数据频率的配对交易中,具有根据数据变化进行自我动态修正的功能.     

开环策略下多阶段均值-方差投资组合优化研究

首先研究开环策略下不同财富动态过程的多阶段均值-方差投资组合优化模型,讨论它们的实际意义和计算方法,其中投资比例财富动态过程模型为高度非线性非凸数学规划.进一步研究投资比例财富动态过程模型实际计算问题,并且通过构造辅助模型,给出投资比例两阶段模型的全局解求解方法并通过数值算例和仿真说明该方法的有效性和准确性.最后通过数值算例比较不同财富动态过程在开环策略下和闭环策略下前沿面的关系,结果表明在闭环策略下三种财富过程等价,但是在开环策略下资产财富模型的前沿面最高、资产调整模型的前沿面次之、投资比例多阶段模型的前沿面最低.     

一种新的组合模型在信用评分中的应用

针对信用评分的特点,首先对传统的Bagging算法进行改进,然后基于不同的基分类器构建多个改进的Bagging组合模型,最后根据一种全新的方式将其组合,得到双层组合模型.通过实证分析得出以下结论:一、与原Bagging算法相比,改进算法能有效降低第二类错误率;二、与单层组合模型相比,多层组合模型能进一步降低第二类错误率和损失率.     

Multi-period portfolio selection problem under uncertain environment with bankruptcy constraint

The complexity of financial markets leads to different types of indeterminate asset returns. For example, asset returns are considered as random variables, when the available data is enough. When the available data is too small or even no available data to estimate a probability distribution, we have to invite some domain experts to evaluate the belief degrees of asset returns. Then, asset returns can be described as uncertain variables. In this paper, we discuss a multi-period portfolio selection problem under uncertain environment, which maximizes the final wealth and minimizes the risk of investment. Unlike the common method to describe the multi-period portfolio selection problem as a bi-objective optimization model, we formulate this uncertain multi-period portfolio selection problem by a new method in three steps with two single objective optimization models. And, we consider the influence of transaction cost and bankruptcy of investor. Then, the proposed uncertain optimization models are transformed into the corresponding crisp optimization models and we use the genetic algorithm combined with penalty function method to solve them. Finally, a numerical example is given to show the effectiveness and practicability of proposed models and method.     

具有熵约束的多阶段均值—绝对偏差模糊投资组合决策

文章运用可能性绝对偏差和比例熵分别度量风险和分散化程度,提出了具有风险控制和线性交易成本的终期财富最大化的多阶段模糊投资组合模型。运用可能理论,将该模型转化为显示的非线性动态优化问题。由于投资过程存在交易成本,上述模型为具有路径依赖性的动态优化问题。文章提出了前向动态规划方法求解。最后, 通过实证研究比较了不同熵的取值投资组合最优投资比例和最终财富的变化。     

Using genetic algorithm to solve a new multi-period stochastic optimization model

This paper presents a new asset allocation model based on the CVaR risk measure and transaction costs. Institutional investors manage their strategic asset mix over time to achieve favorable returns subject to various uncertainties, policy and legal constraints, and other requirements. One may use a multi-period portfolio optimization model in order to determine an optimal asset mix. Recently, an alternative stochastic programming model with simulated paths was proposed by Hibiki [N. Hibiki, A hybrid simulation/tree multi-period stochastic programming model for optimal asset allocation, in: H. Takahashi, (Ed.) The Japanese Association of Financial Econometrics and Engineering, JAFFE Journal (2001) 89-119 (in Japanese); N. Hibiki A hybrid simulation/tree stochastic optimization model for dynamic asset allocation, in: B. Scherer (Ed.), Asset and Liability Management Tools: A Handbook for Best Practice, Risk Books, 2003, pp. 269-294], which was called a hybrid model. However, the transaction costs weren’t considered in that paper. In this paper, we improve Hibiki’s model in the following aspects: (1) The risk measure CVaR is introduced to control the wealth loss risk while maximizing the expected utility; (2) Typical market imperfections such as short sale constraints, proportional transaction costs are considered simultaneously. (3) Applying a genetic algorithm to solve the resulting model is discussed in detail. Numerical results show the suitability and feasibility of our methodology.     

Multi-period portfolio optimization for asset–liability management with bankrupt control

In this paper, we investigate a multi-period portfolio optimization problem for asset–liability management of an investor who intends to control the probability of bankruptcy before reaching the end of an investment horizon. We formulate the problem as a generalized mean–variance model that incorporates bankrupt control over intermediate periods. Based on the Lagrangian multiplier method, the embedding technique, the dynamic programming approach and the Lagrangian duality theory, we propose a method to solve the model. A numerical example is given to demonstrate our method and show the impact of bankrupt control and market parameters on the optimal portfolio strategy.     

基于动态非线性损失厌恶的投资组合优化与实证研究

从行为金融学的角度考虑投资者损失厌恶的心理特征,构建了基于线性损失厌恶和非线性损失厌恶行为投资组合模型。利用中国市场数据模拟一种静态情景和四种动态情景,实证研究不同损失厌恶投资组合模型在不同情景下不同损失厌恶程度的最优资产配置策略和投资绩效表现,并将结果与均值方差模型等传统的投资组合模型进行比较。研究发现损失厌恶投资组合模型优于传统投资组合模型,不同情景下不同程度损失厌恶投资者具有不同的资产配置策略,其投资绩效表现也不尽相同。     

基于综合风险收益的贷款组合优化决策

商业银行货款组合决策的过程,是遵循“效益性、安全性、流动性”的原则,在综合考虑贷款收益和风险的前提下,从众多的贷款对象中选择一组合适的贷款对象的过程。建立综合考虑贷款收益和风险的贷款决策模型,有利于银行通过量化计算进行科学决策,以提高信贷质量,达到商业银行的经营目标。     

基于CVaR衍生的多期多面风险度量下的投资组合研究

为了解决多期投资组合的决策问题,本文将由CVaR衍生的多期多面风险度量作为风险控制目标,建立了一个在收益约束条件下最小化风险的多阶段投资组合模型。为求解模型,设计了多期投资组合优化流程,它将非参数抽样方法、基于聚类算法的多阶段情景树生成方法和多期多面风险度量组合在一起。该流程基于计算、容易实现、直观合理。根据我国金融市场数据进行的实证研究结果表明,这一流程具有较好的实用性。     

Portfolio selection in discrete time with transaction costs and power utility function: a perturbation analysis

In this article, we study a multi-period portfolio selection model in which a generic class of probability distributions is assumed for the returns of the risky asset. An investor with a power utility function rebalances a portfolio comprising a risk-free and risky asset at the beginning of each time period in order to maximize expected utility of terminal wealth. Trading the risky asset incurs a cost that is proportional to the value of the transaction. At each time period, the optimal investment strategy involves buying or selling the risky asset to reach the boundaries of a certain no-transaction region. In the limit of small transaction costs, dynamic programming and perturbation analysis are applied to obtain explicit approximations to the optimal boundaries and optimal value function of the portfolio at each stage of a multi-period investment process of any length.     

A dynamic binomial expansion technique for credit risk measurement: a Bayesian filtering approach

Credit risk measurement and management are important and current issues in the modern finance world from both the theoretical and practical perspectives. There are two major schools of thought for credit risk analysis, namely the structural models based on the asset value model originally proposed by Merton and the intensity‐based reduced form models. One of the popular credit risk models used in practice is the Binomial Expansion Technique (BET) introduced by Moody's. However, its one‐period static nature and the independence assumption for credit entities' defaults are two shortcomings for the use of BET in practical situations. Davis and Lo provided elegant ways to ease the two shortcomings of BET with their default infection and dynamic continuous‐time intensity‐based approaches. This paper first proposes a discrete‐time dynamic extension to the BET in order to incorporate the time‐dependent and time‐varying behaviour of default probabilities for measuring the risk of a credit risky portfolio. In reality, the ‘true’ default probabilities are unobservable to credit analysts and traders. Here, the uncertainties of ‘true’ default probabilities are incorporated in the context of a dynamic Bayesian paradigm. Numerical studies of the proposed model are provided.     

基于双因子CIR强度式定价的信用债券投资组合优化

信用风险和利率风险是相互关联影响的。资产组合优化不能将这两种风险单独考虑或简单的相加,应该进行整体的风险控制,不然会造成投资风险的低估。本文的主要工作:一是在强度式定价模型的框架下,分别利用CIR随机利率模型刻画利率风险因素“无风险利率”和信用风险因素“违约强度”的随机动态变化,衡量在两类风险共同影响下信用债券的市场价值,从而构建CRRA型投资效用函数。以CRRA型投资效用函数最大化作为目标函数,同时控制利率和信用两类风险。弥补了现有研究中仅单独考虑信用风险或利率风险、无法对两种风险进行整体控制的弊端。二是将无风险利率作为影响违约强度的一个因子,利用“无风险利率因子”和“纯信用因子”的双因子CIR模型拟合违约强度,考虑了市场利率变化对于债券违约强度的影响,反映两种风险的相关性。使得投资组合模型中既同时考虑了信用风险和利率风险、又考虑了两种风险的交互影响。避免在优化资产组合时忽略两种风险间相关性、可能造成风险低估的问题。