Adaptação de algoritmos de processamento de dados ambientais para o contexto de Big Data

Abstract

Environmental research depend on sensor generated data to create time series regarding the variables that are being analyzed. The amount of data tends to increase as more and more sensors are created and installed. After some time the datasets become huge and requires new ways to process and store the data. This work seeks to find ways to avoid these issues using a technological solution able to store and process large amounts of data. The solution used is Apache Hadoop, a tool which purpose is to solve Big Data problems. In order to evaluate the tool were used different datasets and time series analysis algorithms. The analysis of chaotic and non-chaotic time series were implemented. These implementations were: the wavelet transform, similarity search using Euclidean distance function, the calculus of the box-counting dimension and the calculus of the correlation dimension. Those implementations were adapted for the MapReduce parallel processing paradigm.