Neural Models for Rainfall Forecasting

This chapter is focused on obtaining an optimal forecast of one month lagged rainfall in Spain. It is assessed by analyzing 22 years of both satellite observations of vegetation activity (e.g. NDVI) and climatic data (precipitation, temperature). The specific influence of non-spatial climatic indices such as NAO and SOI is also addressed. The approaches considered for rainfall forecasting include classical Auto-Regressive Moving-Average with Exogenous Inputs (ARMAX) models and Artificial Neural Networks (ANN), the so-called Multilayer Perceptron (MLP), in particular. The use of neural models is proven to be an adequate mathematical prediction tool in this problem due the non-linearity of the problem. These models enable us to predict, with one month foresight, the general rainfall dynamics, with average errors of 44 mm (RMSE) in a test series of 4 years with a rainfall standard deviation equal to 73 mm. Also, the sensitivity analysis in the neural network models reveals that observations in the status of the vegetation cover in previous months have a predictive power greater than other considered variables. Linear models yield average results of 55 mm (RMSE) although they need a large number of error terms (12) to obtain acceptable models. Nevertheless, they provide means for assessing the seasonal influence of the precipitation regime with the aid of linear dummy regression parameters, thereby offering an immediate interpretation (e.g. coherent maps) of the causality between vegetation cover and rainfall.

Application of Machine Learning Techniques in the Study of the Relevance of Environmental Factors in Prediction of Tropospheric Ozone

This work presents a new approach for one of the main problems in the analysis of atmospheric phenomena, the prediction of atmospheric concentrations of different elements. The proposed methodology is more efficient than other classical approaches and is used in this work to predict tropospheric ozone concentration. The relevance of this problem stems from the fact that excessive ozone concentrations may cause several problems related to public health. Previous research by the authors of this work has shown that the classical approach to this problem (linear models) does not achieve satisfactory results in tropospheric ozone concentration prediction. The authors’ approach is based on Machine Learning (ML) techniques, which include algorithms related to neural networks, fuzzy systems and advanced statistical techniques for data processing. In this work, the authors focus on one of the main ML techniques, namely, neural networks. These models demonstrate their suitability for this problem both in terms of prediction accuracy and information extraction.

Connectionist Systems and Signal Processing Techniques Applied to the Parameterization of Stellar Spectra

A stellar spectrum is the finger-print identification of a particular star, the result of the radiation transport through its atmosphere. The physical conditions in the stellar atmosphere, its effective temperature, surface gravity, and the presence and abundance of chemical elements explain the observed features in the stellar spectra, such as the shape of the overall continuum and the presence and strength of particular lines and bands. The derivation of the atmospheric stellar parameters from a representative sample of stellar spectra collected by ground-based and spatial telescopes is essential when a realistic view of the Galaxy and its components is to be obtained. In the last decade, extensive astronomical surveys recording information of large portions of the sky have become a reality since the development of robotic or semi-automated telescopes. The Gaia satellite is one of the key missions of the European Space Agency (ESA) and its launch is planned for 2011. Gaia will carry out the so-called Galaxy Census by extracting precise information on the nature of its main constituents, including the spectra of objects (Wilkinson, 2005). Traditional methods for the extraction of the fundamental atmospheric stellar parameters (effective temperature (Teff), gravity (log G), metallicity ([Fe/H]), and abundance of alpha elements [a/Fe], elements integer multiples of the mass of the helium nucleus) are time-consuming and unapproachable for a massive survey involving 1 billion objects (about 1% of the Galaxy constituents) such as Gaia. This work presents the results of the authors’ study and shows the feasibility of an automated extraction of the previously mentioned stellar atmospheric parameters from near infrared spectra in the wavelength region of the Gaia Radial Velocity Spectrograph (RVS). The authors’ approach is based on a technique that has already been applied to problems of the non-linear parameterization of signals: artificial neural networks. It breaks ground in the consideration of transformed domains (Fourier and Wavelet Transforms) during the preprocessing stage of the spectral signals in order to select the frequency resolution that is best suited for each atmospheric parameter. The authors have also progressed in estimating the noise (SNR) that blurs the signal on the basis of its power spectrum and the application of noise-dependant algorithms of parameterization. This study has provided additional information that allows them to progress in the development of hybrid systems devoted to the automated classification of stellar spectra.

Soft Computing Techniques in Civil Engineering

Soft computing techniques are applied to a huge quantity of problems spread in several areas of science. In this case, Evolutionary Computation (EC) techniques are applied, in concrete Genetic Programming (GP), to a temporary problem associated to the field of Civil Engineering. The case of study of this technique has been centered in the prediction, over time, of the behavior of the structural concrete in controlled conditions. Given the temporary nature of the case of study, it has been necessary to make several changes to the classical algorithm of GP, among whom it can be emphasized the incorporation of a new operator that gives the GP the ability to be able to solve problems with temporary behavior. The obtained results shown that the proposed method has succeeded in improving the adjustment to the current regulations about creep in the structural concrete.

Electromagnetic Optimization using Genetic Algorithms

This chapter focuses on the application of Genetic Algorithms (GAs) techniques in overcoming the limitations of microstrip antennas in terms of several key parameters such as bandwidth, power-handling capacity etc. In this chapter the effectiveness of GAs is discussed in relation to Electromagnetic optimization. A matching network has been designed for single band and dual band matching of microstrip antenna using GA.

Artificial Intelligence Applied to Natural Resources Management

There are three computational challenges in natural resources management: data management and communication; data analysis; and optimization and control. The authors believe these three challenges can be dealt with Artificial Intelligence (AI) techniques, because they can manage dynamic activities in natural resources. There are several AI techniques such as Genetic Algorithms, Neural Networks, Multi-Agent Systems or Cellular Automata. In this chapter, the authors introduce some applications of Cellular Automata (CA) and Multi-Agent-Based Simulation (MABS) in natural resources management, because these are areas that the authors approach in their research and these areas can contribute to solve the three computational challenges. Specifically, the CA technique can face the challenge of data analysis because it can be extrapolated and new knowledge will be acquired from an area not known or experienced. Regarding the MABS technique, it can solve the challenge of optimization and control, because it works in an empiric way during the decision-making process, based on experiments and observations.

Improving Ontology Alignment through Genetic Algorithms

Ontology alignment is recognized as a fundamental process to achieve an adequate interoperability between people or systems that use different, overlapping ontologies to represent common knowledge. This process consists of finding the semantic relations between different ontologies. There are different techniques conceived to measure the semantic similarity of elements from separate ontologies, which must be adequately combined in order to obtain precise and complete results. Nevertheless, combining multiple measures into a single similarity metric is a complex problem, which has been traditionally solved using weights determined manually by an expert, or calculated through general methods that does not provide optimal results. In this chapter, a genetic algorithm based approach to find out how to aggregate different similarity metrics into a single measure is presented. Starting from an initial population of individuals, each one representing a specific combination of measures, the algorithm finds the combination that provides the best alignment quality.

GA-Based Data Mining Applied to Genetic Data for the Diagnosis of Complex Diseases

A new algorithm is presented for finding genotype-phenotype association rules from data related to complex diseases. The algorithm was based on genetic algorithms, a technique of evolutionary computation. The algorithm was compared to several traditional data mining techniques and it was proved that it obtained better classification scores and found more rules from the data generated artificially. It also obtained similar results when using some UCI Machine Learning datasets. In this chapter it is assumed that several groups of Single Nucleotide Polymorphisms (SNPs) have an impact on the predisposition to develop a complex disease like schizophrenia. It is expected to validate this in a short period of time on real data.

Automatic Arrhythmia Detection

In the present chapter, the authors have developed a tool for the automatic arrhythmias detection, based on time-frequency features and using a Support Vector Machines (SVM) as classifier. Arrhythmia Database Massachusetts Institute of Technology (MIT) has been used in the work in order to detect eight different states, seven are pathologies and one is normal. The unions of different blocks and its optimization have found success rates of 99.82% for RR’ interval detection from electrocardiogram (PQRST waves), and 99.23% for pathologic detection. In particular, the authors have used wavelet transform in order to characterize the wave of electrocardiogram (ECG), based on Biorthogonal family, achieving the most discriminative coefficients. A discussion on arrhythmia ECG classification methods is also presented in this paper.

Intrinsic Evolvable Hardware Structures

The main target of this chapter is to present the intrinsic evolvable hardware structures: concept, design and applications. The intrinsic evolvable hardware structures concept join more research areas like: bio–inspired searching methods (evolutionary algorithms), optimization of algorithms by parallel processing and reconfigurable circuits. First, a general overview about intrinsic evolvable hardware structure is presented. The intrinsic evolvable hardware structure consists of two main modules: hardware genetic algorithm and dynamic reconfigurable circuit. The hardware genetic algorithm searches the configuration that makes the reconfigurable circuit to correctly respond to application requirements. The background section present the genetic algorithm concept as a bio-inspired search solution, the hardware reconfiguration concept with sub areas classifications and the research directions in the evolvable hardware structures areas with application examples. The main section presents the design solutions for hardware implementation of genetic algorithm and for the reconfigurable circuit. Finally, several applications are presented that illustrate the usefulness of the intrinsic evolvable hardware structure.