In this work we present a new hybrid approach for solving the clustering problem for geographic data, which is known to be NP-hard. Two metaheuristics that have proven efficiency in combinatory optimization problems have been chosen for the comparison: Simulated Annealing (SA) and Variable Neighborhood Search (VNS). The proposed model is based on the partitioning around the medoids and on P-median. Previous test runs have shown satisfactory results (in terms of quality and time) for instances of 469 geographic objects, but when instances of greater size are used then variability in the results has been detected. In an effort to achieve better results for the clustering problem, we have incorporated a hybridization of simulated annealing and variable neighborhood search to the geographic clustering problem. We have considered different sizes in the tests runs for distinct groups observing that the solutions obtained with the hybrid approach, named SA-VNS hybrid, overcome SA and VNS when they have been implemented individually. Finally, with the aim of evaluating the benefits of the meta-heuristic proposed, we have measured the internal connection of the obtained clusters by means of the Dunn Index. The results obtained show that the hybrid SA-VNS performs better than SA and VNS with respect to the compactness feature.

No abstract

In the development of this work we describe the design proposal of a useful tool for the monitoring of the GEM detector panels in order to measure the current consumption and in this way to know the useful life of the detector, due to in theory the GEM doesn't produce discharges but in real life there are some small discharges (nano and micro amperes) that deteriorate the detector. In this proposal it is also contemplated to design a software to analyze the interaction of electromagnetic fields within a waveguide that represents the pathways within a printed circuit board using the study of electromagnetic theory in the development of a simulator, which will allow to analyze the effects of these fields in the propagation of the signals that are obtained and to determine if the data obtained are adequate to allow their subsequent analysis.

Resumen. Las redes sociales contienen suficiente elementos para construir corpus acerca de temas novedosos,en dichos corpus existe la posibilidad de que se conviertan en obsoletos debido a la naturaleza efímera de dicha información. La tarea de retroalimentar corpus para mantenerlos vigentes es muy importante y a su vez, una tarea muy difícil cuando se hace manualmente debido a la cantidad de información que se tiene que manejar. En este trabajo se presenta la comparativa de los resultados de la aplicación de la técnica de ensamble de clasificadores Bagging con votación simple para la selección de Tweets candidatos con la finalidad de retroalimentar el corpus de entrenamiento. Dicho corpus está balanceado y dividido en cuatro clases (alegría, tristeza, ira y miedo) y el proceso de clasificación es realizado por medio de tres modelos: Ranking, Naïve Bayes y Probabilidad de Bigramas. La candidatos son seleccionados de un conjunto de prueba etiquetado manualmente y la retroalimentación del corpus será evaluada por medio de pruebas K-Fold Cross Validation.Palabras clave: bagging, ranking, Naïve Bayes, probabilidad de bigramas, retroalimentación de corpus, ensamble de clasificadores.Abstract. The social networks contains enough elements for build corpus about novel topics in which there is the possibility of become obsolete because to their fleeting nature. The feedback corpus task for keeping them current is very important and in turn a difficult and cost task because the used information amount for this. In this work, the comparative among the results of classifier ensemble technique with simple voting in order to candidates Tweets selection for train corpus feedback has been presented. The train corpus is divided in four classes (happiness, sadness, anger and fear) and the classification process is performed by three models: Ranking, Naïve Bayes and Bigrams Probabilities. The candidates are selected from a manually tagged test set and the feedback is evaluated by K-Fold Cross Validation.

This paper presents our approach for SemEval 2016 task 4: Sentiment Analysis in Twitter. We participated in Subtask A: Message Polarity Classification. The aim is to classify Twitter messages into positive, neutral, and negative polarity. We used a lexical resource for pre-processing of social media data and train a neural network model for feature representation. Our resource includes dictionaries of slang words, contractions, abbreviations, and emoticons commonly used in social media. For the classification process, we pass the features obtained in an unsupervised manner into an SVM classifier.