Maria Teresinha Arns Steiner

A Genetic Algorithm for Solving a Capacitated p-Median Problem

Facility-location problems have several applications, such as telecommunications, industrial transportation and distribution. One of the most well-known facility-location problems is the p-median problem. This work addresses an application of the capacitated p-median problem to a real-world problem. We propose a genetic algorithm (GA) to solve the capacitated p-median problem. The proposed GA uses not only conventional genetic operators, but also a new heuristic “hypermutation” operator suggested in this work. The proposed GA is compared with a tabu search algorithm.

Abordagem de um problema médico por meio do processo de KDD com ênfase à análise exploratória dos dados

Knowledge Discovery in Databases - KDD - is a process that consists of several steps, beginning with the collection of data for the problem under analysis and ending with the interpretation and evaluation of the final results. This paper discusses the influence of exploratory data analysis on the performance of Data Mining techniques with respect to the classification of new patterns, based on its application to a medical problem, and compares the performance of these techniques in order to identify the one with the highest percentage of successes. The results of this study lead to the conclusion that, providing this analysis is done properly, it can significantly improve the performance of these techniques and serve as an important tool to optimize the end results. For the problem under study, the techniques involving a Linear Programming model and Neural Networks were the ones showing the lowest percentages of errors for the test sets, presenting good generalization capacities.

Extração de regras de classificação a partir de redes neurais para auxílio à tomada de decisão na concessão de crédito bancário

Credit-risk evaluation is a very important management science problem in the financial analysis area. Neural Networks have received a lot of attention because of their universal approximation property. They have a high predictive accuracy rate, but how they reach their decisions is not easy to understand. In this paper, we present a real-life credit-risk data set and analyzed it using the NeuroRule extraction technique and the software WEKA. The results were considered very satisfactory, reaching more than 80% of accuracy in granting or denying credit on every simulation.

O problema de roteamento no transporte escolar

This paper deals with the vehicle routing problem for school children transportation and describes some Operations Research techniques that can be used to solve it. The problem considers, besides the distances to be traveled by m vehicles, the vehicle availabilities and capacities and the demands in each of the n points. The implementation to a real problem is studied and the results are analised.

The Strategic Decision-Making as a Complex Adaptive System

A company in a competitive environment that wishes to be a benchmark in the business world needs a management model that enables the development of systemic thinking on the part of its executives. In addition to systemic thinking, it is also necessary that executives (i) are aware that the decision-making processes should be shared, (ii) have bounded rationality, and (iii) exert political influence according to their preferences. In this context, the aim of this paper is to describe a conceptual scientific model for strategic decision-making from rules originating from Complex Adaptive Systems and the following mathematical techniques: Analytic Network Process and Linear Programming. This applied and quantitative study is a theoretical essay developed from an integrative review of the aforementioned concepts and techniques, resulting in the proposition of a scientific and conceptual mathematical model that can be applied to a wide variety of business environments. The results obtained from a hypothetical example (Strategic Operation Management Decision) show that the model is able to rank a set of strategic decisions in the environment of most companies and generate information to minimize the negative effects of shared decisions.

Ant Colony Optimisation for Backward Production Scheduling

The main objective of a production scheduling system is to assign tasks (orders or jobs) to resources and sequence them as efficiently and economically (optimised) as possible. Achieving this goal is a difficult task in complex environment where capacity is usually limited. In these scenarios, finding an optimal solution—if possible—demands a large amount of computer time. For this reason, in many cases, a good solution that is quickly found is preferred. In such situations, the use of metaheuristics is an appropriate strategy. In these last two decades, some out-of-the-shelf systems have been developed using such techniques. This paper presents and analyses the development of a shop-floor scheduling system that uses ant colony optimisation (ACO) in a backward scheduling problem in a manufacturing scenario with single-stage processing, parallel resources, and flexible routings. This scenario was found in a large food industry where the corresponding author worked as consultant for more than a year. This work demonstrates the applicability of this artificial intelligence technique. In fact, ACO proved to be as efficient as branch-and-bound, however, executing much faster.

Otimização no serviço de saúde no estado do Paraná: fluxo de pacientes e novas configurações hierárquicas

Neste trabalho e apresentada uma proposta para a otimizacao no servico de saude no estado do Parana com relacao ao fluxo de pacientes dentro do estado e a regionalizacao (divisao) do estado, obtendo novas configuracoes hierarquicas para o mesmo. Quanto a regionalizacao, a proposta consiste em dividir o estado em regioes menores, formadas por varias cidades, vinculadas a uma cidade sede, principal responsavel pelo atendimento no seu nivel de resolutividade. Com relacao ao fluxo de pacientes, e proposto um algoritmo que, ao mesmo tempo em que organiza as informacoes, otimiza o fluxo. Ja para a regionalizacao, fez-se uso do algoritmo branch and price, que utiliza o algoritmo de geracao de colunas em cada no de uma arvore branch and bound. A tecnica proposta apresentada para otimizar o fluxo de pacientes mostrou-se eficaz e util, pois alem de fazer o controle dos procedimentos medicos realizados em cada cidade, tambem define para qual cidade o paciente deve ser encaminhado, respeitando a divisao hierarquica do estado. Ja o algoritmo branch and price, utilizado para a otimizacao na regionalizacao do estado, e bastante interessante, pois tenta melhorar a referida divisao hierarquica do estado, levando em consideracao o numero de habitantes e o numero de procedimentos medicos de cada municipio do estado. Os resultados obtidos tem atendido as expectativas da SESA-PR.

Técnicas da pesquisa operacional no problema de horários de atendentes em centrais telefônicas

Este trabalho propoe uma solucao para a elaboracao e a designacao de jornadas de trabalho em uma central telefonica de atendimento 24 horas. O trabalho foi desenvolvido em tres fases: na primeira, determina-se o numero de atendentes necessarios a cada meia hora do dia por meio de um simulador da central telefonica, visando ao pronto atendimento ao cliente. Na segunda, e determinado o conjunto de jornadas de trabalho que atendem a demanda, minimizando os gastos da empresa com salarios; para isso, foram utilizados os resultados da primeira fase e as jornadas disponiveis para, entao, construir um modelo de Programacao Inteira, o qual foi resolvido com o software LINDO. Finalmente, na terceira fase sao designados os atendentes aos horarios de acordo com suas preferencias, utilizando o algoritmo do Matching de peso maximo. Os resultados encontrados foram analisados em termos de economia para a empresa, melhor atendimento ao usuario e satisfacao dos atendentes.

Técnicas da pesquisa operacional aplicadas na otimização dos serviços postais

O presente trabalho apresenta uma metodologia para a otimizacao do servico de entrega de correspondencias realizado pela Empresa de Correios e Telegrafos (ECT) a partir da aplicacao de algumas tecnicas da Pesquisa Operacional. Como o servico de entrega e as etapas que o antecedem (separacao e ordenacao dos objetos) sao realizadas manualmente, existe a necessidade de otimiza-las e isto e possivel redefindo-se as regioes de atendimento para cada carteiro. Para tanto, o problema foi tratado como um problema de roteamento de veiculos. Varios algoritmos classicos foram utilizados; inicialmente para definir as areas de atendimento dos carteiros, chamados de distritos postais e, a seguir, para estabelecer o roteiro de entrega das correspondencias, considerando, nesta fase, as distâncias reais entre os pontos de entrega. Varios testes computacionais foram realizados, variando-se os algoritmos e parâmetros iniciais e suas respostas comparadas atraves das distâncias totais e de tempos computacionais, determinando-se, assim, os algoritmos com melhores desempenhos.

QUANTUM INSPIRED PARTICLE SWARM COMBINED WITH LIN-KERNIGHAN-HELSGAUN METHOD TO THE TRAVELING SALESMAN PROBLEM

The Traveling Salesman Problem (TSP) is one of the most well-known and studied problems of Operations Research field, more specifically, in the Combinatorial Optimization field. As the TSP is a NP (Non-Deterministic Polynomial time)-hard problem, there are several heuristic methods which have been proposed for the past decades in the attempt to solve it the best possible way. The aim of this work is to introduce and to evaluate the performance of some approaches for achieving optimal solution considering some symmetrical and asymmetrical TSP instances, which were taken from the Traveling Salesman Problem Library (TSPLIB). The analyzed approaches were divided into three methods: (i) Lin-Kernighan-Helsgaun (LKH) algorithm; (ii) LKH with initial tour based on uniform distribution; and (iii) an hybrid proposal combining Particle Swarm Optimization (PSO) with quantum inspired behavior and LKH for local search procedure. The tested algorithms presented promising results in terms of computational cost and solution quality.