Silvia Casado

A variable selection method based on Tabu search for logistic regression models

A Tabu search method is proposed and analysed for selecting variables that are subsequently used in Logistic Regression Models. The aim is to find from among a set of m variables a smaller subset which enables the efficient classification of cases. Reducing dimensionality has some very well-known advantages that are summarized in literature. The specific problem consists in finding, for a small integer value of p, a subset of size p of the original set of variables that yields the greatest percentage of hits in Logistic Regression. The proposed Tabu search method performs a deep search in the solution space that alternates between a basic phase (that uses simple moves) and a diversification phase (to explore regions not previously visited). Testing shows that it obtains significantly better results than the Stepwise, Backward or Forward methods used by classic statistical packages. Some results of applying these methods are presented.

Analysis of new variable selection methods for discriminant analysis

Several methods to select variables that are subsequently used in discriminant analysis are proposed and analysed. The aim is to find from among a set of m variables a smaller subset which enables an efficient classification of cases. Reducing dimensionality has some advantages such as reducing the costs of data acquisition, better understanding of the final classification model, and an increase in the efficiency and efficacy of the model itself. The specific problem consists in finding, for a small integer value of p, the size p subset of original variables that yields the greatest percentage of hits in the discriminant analysis. To solve this problem a series of techniques based on metaheuristic strategies is proposed. After performing some test it is found that they obtain significantly better results than the stepwise, backward or forward methods used by classic statistical packages. The way these methods work is illustrated with several examples.

A tabu search approach to an urban transport problem in northern Spain

In this work we analyze an urban transport problem that the City Council of Burgos, a city in northern Spain, has posed to the authors. Given a fleet of buses and drivers, the problem consists in designing routes and assigning buses to the routes such that the service level is optimized. The optimality of the service level is measured in terms of the waiting time at the bus stops and the duration of the trip. Thus, the problem comprises two decision levels (route design and bus assignment) and differs from other urban transport models found in the literature. In order to solve the problem, we propose two algorithms: one with a local search strategy and another with a tabu search strategy. In both cases, the solutions of the two decision levels are modified in alternating steps. The proposed algorithms obtained significantly better results than the tools currently applied by the transport authorities. In addition, the solutions obtained are very robust with respect to variations on demand, as shown by the experiments.

A computational tool for optimizing the urban public transport: A real application

In this work, we have conducted a study to evaluate and improve the performance of an urban transportation system. Specifically, we have designed an algorithm for obtaining new routes and assigning buses to these routes. The objective is to optimize the service level, measured as the sum of the time the passengers have to wait at the bus stops plus the duration of their journey. As a result, a user-friendly computational tool has been designed, which is currently used by the Burgos City Council. The tool has an attractive graphic interface and is flexible, allowing modifications in the input data. The solutions yielded by the system show an improvement of almost 10% in the service level. The work includes an analysis to identify which set of stops could be “repositioned” to improve even more the service level.

Heuristical labour scheduling to optimize airport passenger flows

We describe the development and implementation of a decision support system for the optimization of passenger flow by trading off service quality and labour costs at an airport. The system integrates a simulation module with an optimization module that requires that Dantzigs labour scheduling problem be solved in the order of thousands of times. We developed a customized scatter search to give the system the capability of finding high-quality solutions to the labour scheduling problems in short computational times. Our experiments verify that our scatter search implementation meets the needed requirements.

Applying genetic algorithms to Wall Street

Genetic algorithms (GAs) can be applied to a wide range of problems in the field of finance. The purpose of this paper is to make GAs accessible to practitioners, academicians and students who are interested in financial markets. By describing a simple application consisting in tuning a technical trading system for the Dow Jones we illustrate step by step how the reader can implement its own trading system with the help of the powerful tool, the GA. To show how this technique can easily be extended to other type of applications in the financial domain, some examples are brought up at the end of the paper.

Variable neighborhood search with memory for a single-machine scheduling problem with periodic maintenance and sequence-dependent set-up times

Abstract In this paper we study the problem of sequencing jobs in a single machine with programmed preventive maintenance and sequence-dependent set-up times. This is an NP-hard problem that has practical relevance because of its industrial applications (textile industry, chemical industry, manufacturing of printed circuit boards, etc.), in which machines need periodic preventive maintenance. An improved formulation of this problem is proposed. Using this new formulation computational experiments show that commercial software can solve exactly not only small-sized instances but also almost all medium-sized instances as well. For solving large-sized instances a heuristic method based on the Variable Neighborhood Search (VNS) is proposed. Specifically, a Skewed VNS with memory, that is, it allows, under certain conditions, the current solution to move to a worse solution and for the incorporation of memory in the search process. Computational experiments show the good performance of our proposed VNS-based method. For small- and medium-sized instances specifically, this method obtains very close-to-optimal solutions, finding the optimal solution in the almost every case. In larger instances our method performs better than previously published algorithms. Several statistical tests support these conclusions. All instances used in computational experiments have been taken from the literature.

A GRASP method for building classification trees

This paper proposes a new method for constructing binary classification trees. The aim is to build simple trees, i.e. trees which are as less complex as possible, thereby facilitating interpretation and favouring the balance between optimization and generalization in the test data sets. The proposed method is based on the metaheuristic strategy known as GRASP in conjunction with optimization tasks. Basically, this method modifies the criterion for selecting the attributes that determine the split in each node. In order to do so, a certain amount of randomisation is incorporated in a controlled way. We compare our method with the traditional method by means of a set of computational experiments. We conclude that the GRASP method (for small levels of randomness) significantly reduces tree complexity without decreasing classification accuracy.