No abstract

Featured Application: A hybrid model formed by neural network techniques and fuzzy systems that predicts rainfall and temperatures in the capital of Minas Gerais, Brazil, which contains several river sources essential for the Brazilian economy. The purpose of the paper is to construct a model capable of extracting knowledge from satellite data related to temperatures and rainfall forecast of the analyzed state.Abstract: The forecast for rainfall and temperatures in underdevelope countries can help in the definition of public and private investment strategies in preventive and corrective nature. Water is an essential element for the economy and living things. This study had a main objective to use an intelligent hybrid model capable of extracting fuzzy rules from a historical series of temperatures and rainfall indices of the state of Minas Gerais in Brazil, more specifically in the capital. Because this is state has several rivers fundamental to the Brazilian economy, this study intended to find knowledge in the data of the problem to help public managers and private investors to act dynamically in the prediction of future temperatures and how they can interfere in the decisions related to the population of the state. The results confirm that the intelligent hybrid model can act with efficiency in the generation of predictions about the temperatures and average rainfall indices, being an efficient tool to predict the water situation in the future of this critical state for Brazil. Appl. Sci. 2019, 9, 5476 2 of 30 watersheds critical nationally. Along with several great cities, it has suffered a strong impact from drought [4].In contrast to the lack of water, it has problems with excess: floods and landslides [2]. If a river or stream receives a considerable amount of water from rain and cannot support it, it ends up overflowing and causing floods, which bring with it destruction, damage, and even death.The facts cited above are just some of the impacts that rain has on society. Within this context, this work was developed for the use of neural networks in the forecast of volumetric rainfall, based on a fuzzy rule system, developing a precise forecast, and contributing to the anticipation of possible tragedies s in places that occur phenomena of greater magnitude, as well as helping in a better control of agriculture. Another purpose was to support air and sea transportation, avoiding accidents in their trajectories. Briefly, when anticipating a possible problem, it is also possible to anticipate solutions and reduce their impacts, such as evacuating areas at risk of landslides or flooding [5].Currently, many kinds of research are being developed to apply models of artificial neural networks in rain forecasts, river basins, and water studies. For example, it was applied to the problem of forecasting the flow of the Nile River in Egypt [6], using a temporal series as a reference to compare tests among several prediction methods of neural networks. Besides, there was a study using stochastic autoregressive mean m...

Resumo: Esse trabalho tem como objetivo realizar um levantamento na literatura sobre as técnicas aplicadas para confecção de arvores de decisão fuzzy e comparar algumas delas com o intuito de analisar sua performance para comprovar a aplicabilidade do método de otimização no cálculo de cobertura de cada regra gerada através de arvores de decisão fuzzy nos mais diversos problemas existentes no mundo real. Os algoritmos abordados neste trabalho são o FID3, FuzzyDT, FCART, FlexDT e uma versão modificada do FuzzyDT nomeada FuzzyDTC. O principal objetivo na utilização de aprendizagem de máquina na verdade é derivar padrões de uma quantidade limitada de dados. Hoje, com o continuo avanço da tecnologia digital, cada vez mais informação é produzida e armazenada. Entretanto, nem todos estes dados são úteis para o aprendizado de máquina, porque neles geralmente estão incluídos juntos aos dados uteis, três outros tipos de dados [3], como os ruidosos, redundantes e incompletos. PalavrasArvores de decisão podem não ter um desempenho satisfatório em uma base de dados grande onde esses três tipos de dados sejam potencializados, dessa maneira a adaptação da lógica fuzzy a esse modelo visa reduzir os impactos da qualidade dos dados [3]. Algoritmos de arvores de decisão fuzzy têm sido propostos a fim de proporcionar uma forma de lidar com incerteza nos dados coletados, já que na literatura encontram-se vários exemplos demonstrando sua superioridade em relação a algoritmos de arvores de decisão crisp como em [4], [5] e [6].

This paper presents a method of data resampling inspired by the operation of a variable selection algorithm using Bayesian techniques. It uses covariance calculations to estimate the minimum mean squared error of the training data and to apply a function to calculate the posterior probability to obtain a more significant number of samples to solve a problem. The model was submitted to standard classification tests, and the results were consistent when compared to other traditional literature models. Resumo: Este artigo apresenta um método de re-amostragem de dados inspirado no funcionamento de um algoritmo de seleção de variáveis através de técnicas Bayesianas. Ele utiliza-se de cálculos da covariância na estimação do erro quadrático médio mínimo dos dados de treinamento e a aplicação de uma função para cálculo da probabilidade posteriori para a obtenção de um maior número de amostras na resolução de um problema. O modelo foi submetido a testes de classificação de padrões e os resultados foram consistentes, quando comparados a outros modelos tradicionais da literatura.