Benchmarking state-of-the-art classification algorithms for credit scoring

In this paper, we study the performance of various state-of-the-art classification algorithms applied to eight real-life credit scoring data sets. Some of the data sets originate from major Benelux and UK financial institutions. Different types of classifiers are evaluated and compared. Besides the well-known classification algorithms (eg logistic regression, discriminant analysis, k-nearest neighbour, neural networks and decision trees), this study also investigates the suitability and performance of some recently proposed, advanced kernel-based classification algorithms such as support vector machines and least-squares support vector machines (LS-SVMs). The performance is assessed using the classification accuracy and the area under the receiver operating characteristic curve. Statistically significant performance differences are identified using the appropriate test statistics. It is found that both the LS-SVM and neural network classifiers yield a very good performance, but also simple classifiers such as logistic regression and linear discriminant analysis perform very well for credit scoring.     

Loading aircraft for military operations

In this paper, we describe an aircraft loading problem submitted by the French military agency (DGA) as part of a more general military airlift planning problem. It can be viewed as a kind of bi-dimensional bin-packing problem, with heterogeneous bins and several additional constraints. We introduce two-phase methods for solving this NP-hard problem. The first phase consists in building good initial solutions, thanks to two fast algorithms: a list-based heuristic and a loading pattern generation method. Both algorithms call a constraint-based subroutine, able to determine quickly if the items already loaded can be reshuffled to accommodate a new object. The second phase improves these preliminary solutions using local search techniques. Results obtained on real data sets are presented.     

Mother-to-child transmission of HIV: a simulation-based approach for the evaluation of intervention strategies

Mother-to-child transmission (MTCT) of HIV is the most significant source of HIV infection in children below the age of 15 years. In 2000 alone, about 600 000 new infections occurred, the vast majority from mothers living in developing countries who were not aware of their HIV-positive status. To date, at least 4.3 million children have died from AIDS. In this paper, we discuss the development of an operational model that can be used to evaluate intervention options for the prevention of MTCT of HIV. The problem was approached by defining suitable pregnancy risk groups, prevention options and the required model outputs to evaluate different intervention strategies. The method chosen to approach the problem was a discrete-event, three-phase simulation, built in Visual Basic, with a stochastic semi-Markov structure. The developed model takes individual pregnant women through each stage of their pregnancy, labour and birth. Different intervention strategies may be modelled at any time, including short-course antiretroviral drugs and cessation of breastfeeding. The model is demonstrated with data from Botswana, which has one of the highest HIV-infection rates in the world.     

Approaches to sharing knowledge in group problem structuring

Problem-structuring techniques are an integral aspect of ‘Soft-OR’. SSM, SAST, Strategic Choice, and JOURNEY Making, all depend for their success on a group developing a shared view of a problem through some form of explicit modelling. The negotiated problem structure becomes the basis for problem resolution. Implicit to this process is an assumption that members of the group share and build their knowledge about the problem domain. This paper explores the extent to which this assumption is reasonable. The research is based on detailed records from the use of JOURNEY Making, where it has used special purpose Group Support software to aid the group problem structuring. This software continuously tracks the contributions of each member of the group and thus the extent to which they appear to be ‘connecting’ and augmenting their own knowledge with that of other members of the group. Software records of problem resolution in real organisational settings are used to explore the sharing of knowledge among senior managers. These explorations suggest a typology of knowledge sharing. The implications of this typology for problem structuring and an agenda for future research are considered.     

A market-based computational approach to collaborative organizational learning

We draw upon the concepts of knowledge market, organizational tacit knowledge, credit assignment, and single-loop learning in proposing a market-based conceptual model for collaborative organizational learning. Our proposed model is characterized by the local competition among seller agents and the global collaboration among winner agents in forming a plan, through a chain of ‘upstream–downstream’ working relationship, for task accomplishment. This feature is achieved through three closely coupled processes: the expert selection process, the capital reallocation process, and the plan formation process. Our model is intended for multiple-step learning environment in which each task consists of a sequence of single-step learning tasks. Learning at the global level is the result of a sequence of nested single-loop learning at the local level.     

Synchronization in supply chains: implications for design and management

Supply chain management literature calls for coordination between the different members of the chain. Materials should be moved from one supplier to the next according to a just-in-time schedule. In this paper, we show that for many supply chain configurations, complete synchronization will result in some members of the chain being ‘losers’ in terms of cost. We develop an algorithm for optimal synchronization of supply chains and provide some guidelines for incentive alignment along the supply chain. In developing the model, we use the economic delivery and scheduling problem model and analyze supply chains dealing with single and multiple components. For single-component supply chains, we derive a closed-form expression for the optimal synchronized cycle time. For multi-component supply chains, we develop an algorithm for finding the optimal synchronized cycle time. We test the performance of the algorithm and show that it provides optimal solutions for a wide range of problems. We illustrate the models with numerical examples.     

When should you sack a football manager? Results from a simple model applied to the English Premiership

What strategy should a football (soccer, in American parlance) club adopt when deciding whether to sack its manager? This paper introduces a simple model assuming that a club's objective is to maximize the number of league points that it scores per season. The club's strategy consists of three choices: the length of the honeymoon period during which it will not consider sacking a new manager, the level of the performance trapdoor below which the manager get the sack, and the weight that it will give to more recent games compared to earlier ones. Some data from the last six seasons of the English Premiership are used to calibrate the model. At this early stage of the research, the best strategy appears to have only a short honeymoon period of eight games (much less than the actual shortest period of 12 games), to set the trapdoor at 0.74 points per game, and to put 47% of the weight on the last five games. A club adopting this strategy would obtain on average 56.8 points per season, compared to a Premiership average of 51.8 points.     

Investigating buyer and seller strategies in online auctions

The annual turnover of online auctions is already in tens of billions of dollars and this amount is predicted to grow substantially over the next few years. Hence, it is important to know how buyers and sellers can influence their chances of success. Therefore, data were collected from eBay auctions for three different categories of collectible items, namely those with a published guide price, those with a rough guide price and those having no easily obtainable guide price. The options available to buyers and sellers of items were then analysed. It was found that it was hard for the seller to influence an item's achieved price significantly, apart from items with no guide price where the starting price could have an effect. Most bidders bid close to the current value and so there were insufficient data to determine the consequences of timing on the placing of high bids. For low bids, delaying a bid was found to improve significantly the chances of winning for one of the data sets.     

Cost–benefit modelling for reliability growth

Decisions during the reliability growth development process of engineering equipment involve trade-offs between cost and risk. However slight, there exists a chance an item of equipment will not function as planned during its specified life. Consequently the producer can incur a financial penalty. To date, reliability growth research has focussed on the development of models to estimate the rate of failure from test data. Such models are used to support decisions about the effectiveness of options to improve reliability. The extension of reliability growth models to incorporate financial costs associated with ‘unreliability’ is much neglected. In this paper, we extend a Bayesian reliability growth model to include cost analysis. The rationale of the stochastic process underpinning the growth model and the cost structures are described. The ways in which this model can be used to support cost–benefit analysis during product development are discussed and illustrated through a simple case.