Abstract

Class imbalance is among the most persistent complications which may confront the traditional supervised learning task in real-world applications. Among the different kind of classification problems that have been studied in the literature, the imbalanced ones, particularly those that represents real-world problems, have attracted the interest of many researchers in recent years. In order to face this problems, different approaches have been used or proposed in the literature, between then, soft computing and ensemble techniques. In this work, ensembles and fuzzy techniques have been applied to real-world traffic datasets in order to study their performance in imbalanced real-world scenarios. KEEL platform is used to carried out this study. The results show that different ensemble techniques obtain the best results in the proposed datasets. © 2018, Springer International Publishing AG, part of Springer Nature.

Abstract

Location based services are limited due to the weakness of the Global Navigation Satellite Systems (GNSS) to work properly in indoor and dense urban environments. Among the different technologies proposed to provide indoor positioning, inertial sensors based pedestrian dead-reckoning (PDR) is one of the more cost-effective solutions thanks to no additional infrastructure is needed to be installed, but its performance is time-limited by drift problems. Regarding the heading drift, some heuristics make use of the buildings' dominant directions in order to reduce this problem. In this paper, the method known as improved heuristic drift elimination (iHDE) is enhanced to be implemented in a Step-and-Heading (SHS) based PDR system, that allows to place the inertial sensors in almost any location of the user's body. Particularly, wrist-worn sensors will be used. Tests on synthetically generated and real data show that the iHDE method can be used in a SHS-based PDR without losing its heading drift reduction capability. © 2016 IEEE.

Abstract

Adaptation to changes in real scenarios is not a 'cost-free' operation. However, in general, this is not considered in most of the studies done in dynamic optimization problems. Our focus here is to analyse what happens when a relocation cost is added in the dynamic maximal covering location problem, that is, when the adaptation to changes entails some cost. Comparing two models with and without 'cost-free' adaptation, respectively, we study how much coverage is lost, how similar are the solutions obtained for both problems along the time and we preliminarily explore the relation between solution similarity and coverage differences. © 2016 IEEE.

Abstract

Location based services can improve the quality of patient care and increase the efficiency of the healthcare systems. Among the different technologies that provide indoor positioning, inertial sensors based pedestrian dead-reckoning (PDR) is one of the more cost-effective solutions, but its performance is limited by drift problems. Regarding the heading drift, some heuristics make use of the building's dominant directions in order to reduce this problem. In this paper, we enhance the method known as improved heuristic drift elimination (iHDE) to be implemented in a Step-and-Heading (SHS) based PDR system, that allows to place the inertial sensors in almost any location of the user's body. Particularly, wrist-worn sensors will be used. Tests on synthetically generated and real data show that the iHDE method can be used in a SHS-based PDR without losing its heading drift reduction capability. © 2016 IEEE.

Abstract

TIMON is an EU research project under the programme Horizon 2020 that aims at creating a cooperative ecosystem integrating traffic information, transport management, ubiquitous data and system self-management. The objective of TIMON is to provide real-time services through a web based platform and a mobile APP for drivers, Vulnerable Road Users (VRUs) and bussinesses. These services will contribute to increasing drivers and VRUs assistance. In this project, many of the services mentioned before are supported by a traffic state prediction system. For this reason, the objective of this study is to lay the groundwork for developing an efficient prediction tool. In this work, a preliminary study is shown, comparing the performance of three different evolutionary methods. © 2016 Copyright held by the owner/author(s).

Abstract

In this paper, a comparative study between a hybrid technique that combines a Genetic Algorithm with a Cross Entropy method to optimize Fuzzy Rule-Based Systems, and literature techniques is presented. These techniques are applied to traffic congestion datasets in order to determine their performance in this area. Different types of datasets have been chosen. The used time horizons are 5, 15 and 30min. Results show that the hybrid technique improves those results obtained by the techniques of the state of the art. In this way, the performed experimentation shows the competitiveness of the proposal in this area of application. © Springer International Publishing Switzerland 2016.

Abstract

Accurate estimation of the future state of the traffic is an attracting area for researchers in the field of Intelligent Transportation Systems (ITS). This kind of predictions can lead to traffic managers and drivers to act in consequence, reducing the economic and social impact of a possible congestion. Due to the inter-urban traffic information nature, the task of predicting the future state of the traffic requires, in most cases, a non-linear patterns search in the input data. In recent years, a wide variety of models has been used to solve this problem in the most accurate way. Due to that, models generated to provide information about the future state of the road are, usually, incomprehensible to a human operator, making impossible to give him/her an explanation about the causes of the prediction. Given the capacity of rule based systems to explain the reasoning followed to classify a new pattern, the advantages and disadvantages of such approaches are explored in this work. To conduct such task, datasets recorded from the California Department of Transportation are created. A 9-kilometer section of the I5 highway of Sacramento is used for this research. Two different types of datasets are built for the experimentation. One of them contains the entire information recorded. The other one contains with a simplified version of the information, considering only the first, middle and last monitored points of the road. Twelve prediction horizons, from 5 to 60 minutes, were considered for prediction. An experimental comparative study involving 16 state of the art techniques is performed. Techniques tested include those that fall within the categories of Evolutionary Crisp Rule Learning (ECRL) and Evolutionary Fuzzy Rule Learning (EFRL). These methods were selected since they offer to the final user, not only a prediction, but also a legible model about the way in which the decision was taken. Techniques are compared in terms of accuracy and complexity of the models generated. © 2016 The Authors.

Abstract

This paper studies how the accuracy of the step detection algorithm of a pedestrian dead-reckoning (PDR) system is affected by the sampling frequency and the filtering of the data gathered from a wrist-worn inertial measurement unit (IMU). On the one hand, results show that sensors sampling rate can be reduced and a similar accuracy can still be obtained, what it is very interesting for energy saving purposes. However, a low sampling frequency requires a finer tuning of the algorithm's parameters. On the other hand, the application of a filter to the data gathered from the sensors is always recommended, in order to get some performance improvement. Different types of filters can be used in function of the value of the sampling frequency. © 2015 IEEE.

Abstract

In this paper, a metaheuristic that combines a Genetic Algorithm and a Cross Entropy Algorithm is presented. The aim of this work is to achieve a synergy between the capabilities of the algorithms using different population sizes in order to obtain the closest value to the optimal of the function. The proposal is applied to 12 benchmark functions with different characteristics, using different configurations.

Abstract

La gestión de la incertidumbre en los problemas de localización de cobertura máxima (PLCM) es muy importante debido a la naturaleza imprecisa de algunos elementos de estos problemas en el mundo real. La demanda generada en los nodos, la distancia entre nodos, la disponibilidad del servicio y el radio de cobertura son los parámetros que comúnmente se pueden considerar inciertos. Los enfoques probabilísticos y los enfoques difusos han sido los más utilizados para modelar la incertidumbre en el PLCM, pero los de corte difuso no están estructurados o sistematizados. Por eso en este artículo se presenta una revisión de los trabajos que han abordado el PLCM con incertidumbre de tipo difuso, como paso previo para el planteamiento sistemático de modelos y soluciones para el mismo.

Abstract

Una pregunta que surge en el diseño de esquemas de aprendizaje para Problemas Dinámicos de Optimización consiste en decidir qué hacer con el conocimiento que se ha adquirido una vez que se produce un cambio: olvidarlo o utilizarlo en los cambios posteriores. En este trabajo, intentamos arrojar a la luz sobre este asunto usando el método Adaptive Hill Climbing Memetic Algorithm y el problema de la Mochila como banco de pruebas. Los resultados obtenidos muestran que la respuesta a la pregunta planteada depende de la estructura de la instancia y justifica la necesidad de realizar estudios más completos.

Abstract

En el mundo moderno, las grandes empresas necesitan predecir los cambios en el mercado y anticiparse a ellos tomando las decisiones más adecuadas en el presente. Entre las herramientas especializadas para ello se encuentra la Planificación de Escenarios. Una de las técnicas usadas en Planificación de Escenarios es el Análisis Morfológico, cuya finalidad es la generación sistemática y evaluación de todas las posibles combinaciones de valores que las variables objeto de estudio pueden tomar, descartando aquellas incompatibles en la práctica. Cada combinación representa un posible escenario futuro, es decir, una situación plausible que puede llegar a darse. La variante conocida como MORPHOL evalúa un escenario mediante la probabilidad de que ocurra finalmente. Aquí se revisa una mejora publicada recientemente que considera otros criterios distintos a la probabilidad para evaluar un escenario usando valores lingüísticos. Además, se proponen mejoras futuras y nuevas líneas que están actualmente siendo investigadas, como clustering de los escenarios generados y obtención de descripciones lingüísticas de cada escenario, con ejemplos sobre un caso de estudio real.

Abstract

Companies that want to be competitive must make use of good practices to anticipate the future by analyzing the possible effects of today's decisions on their own long-term development. Scenario planning is among the most extended approaches to accomplish this. One of the techniques often used in scenario planning is Morphological Analysis, which aims to explore the space of feasible futures in a systematic way by analyzing all the combinations of the possible states of the variable that compose the system under study. Each of this combinations represent a possible future scenario. This work focuses on a particular variant known as the MORPHOL method, in which every scenario is evaluated in terms of the probability it eventually arises. This is computed using the marginal probability estimates of the hypothetical variables' states involved in the scenario, which are given by human experts. This method presents two drawbacks: first, the probabilities have to be expressed by numerical values which makes difficult its estimation by humans and does not capture its uncertainty; and second, it examines the scenarios basing only on their probability, thus it may ignore scenarios that are interesting but not the very probable. In order to ease the experts' task and capture their opinions in a better way, we introduce Computing with Words techniques. For solving the second shortcoming, we apply Multi-criteria Decision Making to uncover good scenarios according to several criteria jointly including probability. The result is a novel linguistic multi-criteria method for morphological analysis that has been successfully applied to a real problem and thus deserves further research.

Abstract

La prospectiva tecnológica puede definirse como un conjunto de estudios que se llevan a cabo con el fin de anticipar cuál va a ser el futuro en una determinada área. Una metodología ampliamente utilizada para ello es el Método de los Escenarios de Godet, que incluye un módulo para realizar el conocido como análisis estructural (MIC-MAC). Su finalidad es determinar las variables más importantes en el sistema que se va a estudiar, tomando como entradas las relaciones de influencia existentes entre ellas, las cuales vienen cuantificadas numéricamente por un grupo de expertos en forma de matriz. Proponemos una extensión difusa de este método que permite a los expertos juzgar las relaciones mediante etiquetas lingüísticas y obtener resultados más fácilmente intepretables. Esto es una importante ventaja sobre el método tradicional, ya que en este tipo de estudios, la interpretabilidad de las salidas y los resultados intermedios es un aspecto primordial. Se ha realizado una implementación en Java de nuestro procedimiento difuso con una interfaz web accesible públicamente, y que demuestra la facilidad de uso del enfoque planteado. Los primeros resultados son interesantes desde el punto de vista práctico.

Abstract

Developing predictive models is one of the key issues in Systems Biology. A critical problems that arises when these models are built is the parameter estimation. The calibration of these nonlinear dynamic models is stated as a nonlinear programming problems (NLP) and its resolution is usually complex due to the frequent ill-conditioning and multimodality of the majority of these problems. For that reason, the use of hybrid stochastic optimization methods has received an increasing interest in recent years. In this work we present a new hybrid method for parameter estimation in Systems Biology. This proposal consists on a set of DE algorithms that cooperate among them through a centralised scheme in which a coordinator controls their behavior by means of a rule system. The comparison with state-of-the-art methods shows the better performance of this cooperative strategy when the complexity of the instances is increased.

Abstract

Technology foresight deals with the necessity of anticipating the future to better adapt to new situations regarding innovations that directly affect business world. One widely spread methodology in technology foresight is Godet's Scenario Method, which includes a module (MICMAC) performing the so-called structural analysis. The goal of the structural analysis is to identify the most important variables in a system. To this end, it makes use of an influence matrix that describes the relations between the variables. This information is usually given by experts based on their own knowledge and experience. However, some of the information of the influence matrix may contain errors due to the subjective nature of the criteria and opinions of the experts. Here we propose a new analysis that follows a multi-objective approach and allows to measure the sensibility of the model versus possible errors at the input. The well-known NSGA-II algorithm has been used as a solver. The results are encouraging and deserve further investigation.

Abstract

Dentro de los distintos tipos de metaheurísticas híbridas, las estrategias cooperativas se presentan como una de las alternativas más prometedoras. Este tipo de métodos consisten en un conjunto de algoritmos que exploran el espacio de búsqueda de forma simultánea mientras intercambian información entre ellos. Con el objetivo de profundizar en el estudio de dichos métodos y utilizando el problema del p-hub mediano, en este trabajo analizamos como afecta al comportamiento de la estrategia cooperativa el hecho de cambiar de forma notoria la configuración de las metaheurísticas que la componen.

Abstract

Optimisation in dynamic environments is a very active and important area which tackles problems that change with time (as most real-world problems do). The possibility to use a new centralised cooperative strategy based on trajectory methods (tabu search) for solving Dynamic Optimisation Problems (DOPs) was previously introduced showing good results against state of the art methods like the Particle Swarm Optimisation (PSO) variant with multiple swarms and different types of particles. The analysis of this previous work are further extended here by exploring more possibilities for the cooperation rules used in the strategy. The results show that different classes of cooperation can lead to quite different results, some of them greatly outperforming the previous ones.

Abstract

Metaheuristics are excellent tools for addressing hard combinatorial optimization problems. Given a limited quantity of time and space resources, heuristic algo- rithms are very effective in providing good quality solutions. However, among the open problems in this research field, we want to the outline following ones:

• given a problem, it is sometime impossible to determine a priori which is the best heuristics to solve it
• the performance of a metaheuristics depends dramatically on certain parameters whose adjustment is not trivial, and is usually done by experts or extensive experimentation

The first issue can be handled by cooperative strategies [1], where a set of potentially good heuristics for the optimization problem are executed in parallel, sharing information during the run. The second problem is successfully addressed by Reactive strategies [2], and the use of sub-symbolic machine learning to automate the parameter tuning process, making it an integral part of the algorithm. In [3], a centralized cooperative strategy is presented where a coordinator controls a set of heuristics by means of a Fuzzy rule base. The aim of this paper is to investigate the effectiveness of a coordinator driven by Reactive rules. An empirical comparison between both rules is provided using the Uncapacitated Single Allocation p-Hub Median Problem (USApHMP) [4] as example.

Abstract

La mayoría de las metaheurísticas con mecanismos de auto-adaptación que se han propuesto en la literatura, afrontan cada resolución desde el principio, sin tener en cuenta lo aprendido durante otras búsquedas. Basándonos en la idea de que el esfuerzo computacional realizado para la resolución de una instancia debería ser aprovechado en la solución de instancias similares, presentamos un nuevo modelo que permite la solución simultánea de un grupo de instancias. Este modelo contiene un conjunto de agentes que operan sobre un espacio de búsqueda. Presenta un esquema multi-nivel de 2 capas que controla jerárquicamente la autoadaptación de los parámetros de los operadores. Este nuevo método ha sido aplicado al problema MAX-SAT con conjuntos de varias instancias obteniéndose unos resultados prometedores.