Modelling LGD for unsecured retail loans using Bayesian methods

Loss Given Default (LGD) is the loss borne by the bank when a customer defaults on a loan. LGD for unsecured retail loans is often found difficult to model. In the frequentist (non-Bayesian) two-step approach, two separate regression models are estimated independently, which can be considered potentially problematic when trying to combine them to make predictions about LGD. The result is a point estimate of LGD for each loan. Alternatively, LGD can be modelled using Bayesian methods. In the Bayesian framework, one can build a single, hierarchical model instead of two separate ones, which makes this a more coherent approach. In this paper, Bayesian methods as well as the frequentist approach are applied to the data on personal loans provided by a large UK bank. As expected, the posterior means of parameters that have been produced using Bayesian methods are very similar to the frequentist estimates. The most important advantage of the Bayesian model is that it generates an individual predictive distribution of LGD for each loan. Potential applications of such distributions include the downturn LGD and the stressed LGD under Basel II.     

Dynamic project selection and funding under risk: A decision tree based MILP approach

This paper presents a mixed integer linear programming (MILP) model for a new class of dynamic project selection and funding problems under risk given multiple scarce resources of different qualifications. The underlying stochastic decision tree concept extends classical approaches mainly in that it adds a novel node type that allows for the continuous control of discrete branching probability distributions. The control functions are piecewise linear and are convex for the costs and concave for the benefits. The MILP-model has been embedded in a prototype Decision Support System (DSS). With respect to the proposed solution the DSS provides complete probability distributions for both costs and benefits.     

Research in decision theory: A personal perspective

This essay offers a personal view of developments in two main areas of decision theory — decision making under risk and uncertainty, and social choice theory — with emphasis on the past half century and on likely future directions. Remarks on helpful research practices are also included.     

TASC: Two-attribute-set clustering through decision tree construction

Clustering is the process of grouping a set of objects into classes of similar objects. In the past, clustering algorithms had a common problem that they use only one set of attributes for both partitioning the data space and measuring the similarity between objects. This problem has limited the use of the existing algorithms on some practical situation. Hence, this paper introduces a new clustering algorithm, which partitions data space by constructing a decision tree using one attribute set, and measures the degree of similarity using another. Three different partitioning methods are presented. The algorithm is explained with illustration. The performance and accuracy of the four partitioning methods are evaluated and compared.     

Interactive decision making for uncertain minimum spanning tree problems with total importance based on a risk-management approach

This paper deals with a minimum spanning tree problem where each edge cost includes uncertainty and importance measure. In risk management to avoid adverse impacts derived from uncertainty, a d-confidence interval for the total cost derived from robustness is introduced. Then, by maximizing the considerable region as well as minimizing the cost-importance ratio, a biobjective minimum spanning tree problem is proposed. Furthermore, in order to satisfy the objects of the decision maker and to solve the proposed model in mathematical programming, fuzzy goals for the objects are introduced as satisfaction functions, and an exact solution algorithm is developed using interactive decision making and deterministic equivalent transformations. Numerical examples are provided to compare our proposed model with some previous models.     

A problem-solving approach to product design using decision tree induction based on intuitionistic fuzzy

Customer complaint problem is a product design used to understand customer requirements. Furthermore, product design corresponding to customer requirement does not feel adequately solved for a cause of problem. The cause of the problem affecting product design is solved to prevent customer complaint from reoccurring. However, the problems by customer may have observation uncertainty and fuzzy. Fuzzy concept considers not only the degree of membership to an accept set, but also the degree of non-membership to a rejection set. Therefore, we present a new approach for problem solving using decision tree induction based on intuitionistic fuzzy sets in this paper. Under this approach, we first develop the problem formulation for the symptoms and causes of the problem based on intuitionistic fuzzy sets. Next, we identify the cause of the problem using intuitionistic fuzzy decision tree by the problem formulation. We then provide the approach to find the optimal cause of the problem for the consideration of product design. A numerical example is used to illustrate the approach applied for product design.     

Modelling of hydrological persistence for hidden state Markov decision processes

A reservoir in south east Queensland can supply irrigators, industry or domestic users. Stochastic inflow is modelled by a hidden state Markov chain, with three hidden states corresponding to prevailing climatic conditions. A stochastic dynamic program that relies on estimation of the hidden state is implemented. The optimal decisions are compared with those obtained if the hidden state Markov chain model is replaced with a model that relies on the Southern Oscillation Index to define prevailing climatic conditions.     

Modelling lifetime data: a graphical approach

In this paper we point out the differences between the most common hazard-based models, such as the proportional hazards and the accelerated failure time models. We focus on the heteroscestaticity-across-individuals problem that cannot be accommodated by them, and give motivation and general ideas about more flexible formulations. We describe hybrid and extended models, which have the former models as particular cases, but keep enough flexibility to fit data with heteroscedasticity. We show that by considering simple graphical procedures it is easy to verify whether there is heteroscedasticity in the data, whether it is possible to describe it through a simple function of the covariates, and whether it is important to take it in account for the final fit. Real datasets are considered. Copyright © 1999 John Wiley & Sons, Ltd.     

An adaptive routing approach for personal rapid transit

Personal Rapid Transit (PRT) is a public transportation mode, in which small automated vehicles transport passengers on demand. Central control of the vehicles leads to interesting possibilities for optimized routings. The complexity of the involved routing problems together with the fact that routing algorithms for PRT essentially have to run in real-time often leads to the choice of fast greedy approaches. The most common routing approach is arguably a sequential one, where upcoming requests are greedily served in a quickest way without interfering with previously routed vehicles. The simplicity of this approach stems from the fact that a chosen route is never changed later. This is as well the main drawback of it, potentially leading to large detours. It is natural to ask how much one could gain by using a more adaptive routing strategy. This question is the main motivation of this article. In this paper, we first suggest a simple mathematical model for PRT, and then introduce a new adaptive routing algorithm that repeatedly uses solutions to an LP as a guide to route vehicles. Our routing approach incorporates new requests in the LP as soon as they appear, and reoptimizes the routing of all currently used vehicles, contrary to sequential routing. We provide preliminary computational results that give first evidence of the potential gains of an adaptive routing strategy, as used in our algorithm.     

Modelling stochastic decision systems using dependent-chance programming

This paper further discusses the techniques of dependent-chance programming, dependent-chance multiobjective programming and dependent-chance goal programming. Some illustrative examples are provided to show how to model complex stochastic decision systems by using dependent-chance programming and how to solve these models by employing a Monte Carlo simulation based genetic algorithm.     

An interactive approach for multiobjective decision making

We develop an interactive approach for multiobjective decision-making problems, where the solution space is defined by a set of constraints. We first reduce the solution space by eliminating some undesirable regions. We generate solutions (partition ideals) that dominate portions of the efficient frontier and the decision maker (DM) compares these with feasible solutions. Whenever the decision maker prefers a feasible solution, we eliminate the region dominated by the partition ideal. We then employ an interactive search method on the reduced solution space to help the DM further converge toward a highly preferred solution. We demonstrate our approach and discuss some variations.     

Genetic algorithms in logic tree decision modeling

An important approach to decision modeling is the induction of knowledge structures—such as rules, trees, and graphs—from empirical data describing previous conditions and the resulting decisions. We examine here a specific knowledge structure, a logic tree, in which the conditions are leaves, the decision is the root, and the intermediate nodes are logical operators. We then use genetic algorithms (GAs) to construct logic trees that best represent the correspondence between conditions and decisions described by the empirical data. We also investigate an important characteristic of the GA search, the fitness distance correlation. Finally, we comment on the usefulness of GAs in knowledge modeling.     

Modelling enterprise networks: A master equation approach

We are interested in the problem of the mathematical modelling and simulation of groups of small enterprises. This is where a network of small enterprises is established in order to achieve a global objective. Each enterprise contributes to the network but keeps its own individual identity. Apart from having to model economic factors, we have to consider social phenomena and human factors which have an important influence on the dynamical evolution of the network. This paper is devoted to an approach for modelling the dynamical evolution of the social phenomena, based on a master equation approach and introducing non-linear effects.     

Improved bounds for the randomized decision tree Complexity of recursive majority

We consider the randomized decision tree complexity of the recursive 3‐majority function. We prove a lower bound of for the two‐sided‐error randomized decision tree complexity of evaluating height h formulae with error . This improves the lower bound of given by Jayram, Kumar, and Sivakumar (STOC'03), and the one of given by Leonardos (ICALP'13). Second, we improve the upper bound by giving a new zero‐error randomized decision tree algorithm that has complexity at most . The previous best known algorithm achieved complexity . The new lower bound follows from a better analysis of the base case of the recursion of Jayram et al. The new algorithm uses a novel “interleaving” of two recursive algorithms. © 2015 Wiley Periodicals, Inc. Random Struct. Alg., 48, 612–638, 2016     

LTC: A latent tree approach to classification

Latent tree models were proposed as a class of models for unsupervised learning, and have been applied to various problems such as clustering and density estimation. In this paper, we study the usefulness of latent tree models in another paradigm, namely supervised learning. We propose a novel generative classifier called latent tree classifier (LTC). An LTC represents each class-conditional distribution of attributes using a latent tree model, and uses Bayes rule to make prediction. Latent tree models can capture complex relationship among attributes. Therefore, LTC is able to approximate the true distribution behind data well and thus achieves good classification accuracy. We present an algorithm for learning LTC and empirically evaluate it on an extensive collection of UCI data. The results show that LTC compares favorably to the state-of-the-art in terms of classification accuracy. We also demonstrate that LTC can reveal underlying concepts and discover interesting subgroups within each class.     

A general decomposition approach for multi-criteria decision trees

In this paper, we address the dynamic and multi-criteria decision-making problems under uncertainty, generally represented by multi-criteria decision trees. Decision-making consists of choosing, at each period, a decision that maximizes the decision-maker outcomes. These outcomes should often be measured against a set of heterogeneous and conflicting criteria. Generating the set of non-dominated solutions is a common approach considered in the literature to solve the multi-criteria decision trees, but it becomes very challenging for large problems. We propose a new approach to solve multi-criteria decision trees without generating the set of all non-dominated solutions, which should reduce the computation time and the cardinality of the solution set. In particular, the proposed approach combines the advantages of decomposition with the application of multi-criteria decision aid (MCDA) methods at each decision node. A generalization of the Bellman’s principle of decomposition to the multi-criteria context is put forward. A decomposition theorem is therefore proposed. Under the sufficient conditions stated by the theorem, the principle of decomposition will generate the set of best compromise strategies. Seven MCDA methods are then characterized (lexicographic, weighted sum, multi-attribute value theory, TOPSIS, ELECTRE III, and PROMETHEE II) against the conditions of the theorem of decomposition and against other properties (neutrality, anonymous, fidelity, dominance, independency), in order to confirm or infirm their applicability with the proposed decomposition principle. Moreover, the relationship between independency and temporal consistence is discussed as well as the effects of incomparableness, rank reversals, and use of thresholds. Two conjectures resulted from this characterization.     

Approximate receding horizon approach for Markov decision processes: average reward case

We consider an approximation scheme for solving Markov decision processes (MDPs) with countable state space, finite action space, and bounded rewards that uses an approximate solution of a fixed finite-horizon sub-MDP of a given infinite-horizon MDP to create a stationary policy, which we call “approximate receding horizon control.” We first analyze the performance of the approximate receding horizon control for infinite-horizon average reward under an ergodicity assumption, which also generalizes the result obtained by White (J. Oper. Res. Soc. 33 (1982) 253-259). We then study two examples of the approximate receding horizon control via lower bounds to the exact solution to the sub-MDP. The first control policy is based on a finite-horizon approximation of Howard's policy improvement of a single policy and the second policy is based on a generalization of the single policy improvement for multiple policies. Along the study, we also provide a simple alternative proof on the policy improvement for countable state space. We finally discuss practical implementations of these schemes via simulation.     

Reasoning maps for decision aid: an integrated approach for problem-structuring and multi-criteria evaluation

This paper proposes a tool for multi-criteria decision aid to be referred to as a Reasoning Map. It is motivated by a desire to provide an integrated approach to problem structuring and evaluation, and in particular, to make the transition between these two processes a natural and seamless progression. The approach has two phases. In the first one, the building of a Reasoning Map supports problem structuring, capturing a decision maker's reasoning as a network of means and ends concepts. In the second phase, this map is enhanced, employing a user-defined qualitative scale to measure both performances of decision options and strengths of influence for each means–end link. This latter phase supports the decision maker in evaluating the positive and negative impacts of an action through synthesis of the qualitative information. A case study, which investigates the use of the method in practice, is also presented.