Mercedes Valdés

Approximative fuzzy rules approaches for classification with hybrid-GA techniques

Abstract In this paper the use of different methods from the fuzzy modeling field for classification tasks is evaluated and the potential of their integration in producing better classification results is investigated. The methods considered, approximative in their nature, consider different integrations of techniques with an initial rule generation step and a following rule tuning approach using different evolutionary algorithms. We analyse the adaptation of existing techniques in the fuzzy modeling context for the classification problem, and the integration of these techniques in order to improve the classifiers performance. Finally a genetic algorithm (GA) for translation from approximative rules to similar descriptive ones trying to preserve the accuracy of the approximative classifier is presented. The classical Iris and Cancer data set are used throughout the evaluation process to form a common ground for comparison and performance analysis.

METALA: A Meta-learning Architecture

Meta-learning has been accepted, in the last five years, as a proper machine learning research field. In this concrete area of interest, the way in which different theories, each one produced either with the same algorithm or with many of them, are merged to produce a more accurate model has been the main topic. Now, new emerging techniques got more to do with inductive meta-learning. It is the process of learning from others learning experiences. This kind of learning imposes severe requisites, from the point of view of the software system that would support it. The purpose of this work is to show a software architecture for this type of learning. The architecture will give recommendations for building a system of this kind, that has to tackle with very precise but difficult problems at a time.

An Approach for Representing Sensor Data to Validate Alerts in Ambient Assisted Living

The mainstream of research in Ambient Assisted Living (AAL) is devoted to developing intelligent systems for processing the data collected through artificial sensing. Besides, there are other elements that must be considered to foster the adoption of AAL solutions in real environments. In this paper we focus on the problem of designing interfaces among caregivers and AAL systems. We present an alert management tool that supports carers in their task of validating alarms raised by the system. It generates text-based explanations—obtained through an argumentation process—of the causes leading to alarm activation along with graphical sensor information and 3D models, thus offering complementary types of information. Moreover, a guideline to use the tool when validating alerts is also provided. Finally, the functionality of the proposed tool is demonstrated through two real cases of alert.

Toward a framework for the specification of hybrid fuzzy modeling

During the few last years, several successful approaches for the integration of soft computing techniques have been proposed in the area of data‐driven fuzzy modeling (DDFM). However, there is a lack of methodological and general purpose hybridization in an easy and unified manner. This work outlines the design of a new DDFM framework called METHOD, offering the functionalities needed to combine techniques into hybrid strategies for DDFM tasks. Bearing in mind our main goal, a previous analysis of several existing DDFM techniques helps us: (1) to identify the most usual interaction schemes, by means of which methods are combined into DDFM hybrid strategies; (2) to exemplify requirements and effects for different techniques determining suitable combinations; and (3) to establish the universe of discourse based on which of these requirements and effect are defined. All these ideas are illustrated with examples.

Toward a framework for the specification of hybrid fuzzy modeling: Research Articles

The spread spectrum modulation presents a more robust interference rejection capability than non-code–based modulations. However, the increase of the communication services has augmented the power of the interference signals. One example is unlicensed bands such as the 2.4–2.5 GHz one, where the DS-CDMA communications of the IEEE 802.11b standard are interfered with by other Industrial Scientific and Medical (ISM) services. One way of improving the interference robustness of the DS-CDMA receivers is introducing an interference canceler (IC) before the despreading stage. The optimum criterion to estimate or detect the interference is the maximum-a-posteriori (MAP) one. However, it requires a high computational burden and a complete knowledge of the statistical model, which reduce its feasibility. To overcome these drawbacks this article proposes an interference canceler based on Fuzzy Logic that allows (1) suppressing both analog and digital interferences, and (2) easily controlling the complexity versus performance trade-off. In other words, the design of the fuzzy canceler follows a down/top strategy; that is, a first low complex design is possible just departing from linguistic information and, in a second stage, we refine it, incorporating the available statistical information into its structure. Finally, to assess the proposed fuzzy canceler we have derived asymptotic performance bounds.

Some Applications of Hybrid Fuzzy Modeling

Fuzzy modeling is an effective approach for system identification. It is based on fuzzy sets and logic and describes the system behaviour by means of fuzzy IF-THEN rules. In its turn, data driven fuzzy modeling (DDFM) extracts these models from a set of input-output observations about the system. Three main stages compose DDFM: rules number identification, rules generation and parameter optimization. One way to carry out a DDFM process is by means of a combination of techniques, each one solving one of the DDFM phases. In this paper, the authors applied hybridizations of clustering algorithms and neural networks (NN) in order to solve several regression problems from different domains showing up the suitability and success of hybridization in DDFM

Neuro-Fuzzy Modeling Applied to GIS: a Case Study for Solar Radiation

Soft computing can be a powerful approach in the agricultural domain, from two different and interesting perspectives. The first one is using this to obtain fuzzy models of the environment, in order to create an approximated theory to explain a set of observations which reect a concrete agricultural process. The second is the integration of neuro-fuzzy models in a Geographical Information System, in where spatial and temporal interpolation is a central issue. In this paper we present the use we have done of soft computing for both mentioned perspectives. In particular, here we model irrigation water needs for a concrete zone of Spain and show how the neuro-fuzzy model can be integrated inside a GIS in order to deploy such approximative power.

Data Mining Applied to Irrigation Water Management

This work addresses the application of data mining to obtain artificial neural network based models for the application in water management during crops irrigation. This problem is very important in the zone of the South-East of Spain, as there is an important lack of rainfall there. These intelligent analysis techniques are used in order to optimize the consumption of such an appreciated and limited resource.

Towards a modeling framework for integrating hybrid techniques

Nowadays, it is easy to find a number of different hybrid approaches for fuzzy modeling. All these approaches were built in a very ad-hoc manner, and did not follow a systematic approach. However, we think that some kind of information system which helps in the study of how algorithms can combine to model systems in a fuzzy fashion should be very helpful. In this article, we propose METALA (META-Learning Architecture), an architecture to study the typical processes of machine learning, to study the particular issue of fuzzy modeling.

Fuzzy and hybrid methods applied to GIS interpolation

Soft computing can be a powerful approach in the agricultural domain. In particular two different techniques have been developed for the approximation of solar radiation, a key parameter for the calculus of irrigation water for crops. Besides, one of the methods has been developed on the basis of the cooperation of two well known techniques, and it is presented as a novel approach.