Predição de séries temporais climáticas com aprendizagem profunda

Abstract

The ability to predict climatic conditions is fundamental for decision-making in different areas, such as agriculture, livestock, construction, maritime transport, aeronautics and tourism, as well as allowing the execution of human nature actions. Technological advancement, with the development of sensors, computational resources and approaches to understanding climate variations, comes with the application of predetermined weather models. In general, to build these models, you need climate time series, which is data collected about one or more variables over time. From time series, model models are estimated to produce predictions about variables based on historical data. Gls AP Models allow automatic extraction of resources directly from data, allowing the construction of functions capable of modeling patterns. Therefore, in this paper, it offers an application of gls AP models to build climate data editors. To validate the results, models can use real data, these measures are used to validate a modeling and a solution.