Creation of Time Series Models based on Deep Learning for Generation of Probabilistic Forecasts Using GluonTS and Python

Time series have a wide field of application, mainly e-commerce, finance, medicine, and biology; also, time series can be applied to other specific fields such as forecast and production systems. This disclosure article describes the use of the GluonTS library for the creation of probabilistic time series based on deep learning models to make forecasts related to the production of crude oil by a federal entity, considering as a practical example the set of open data published by the Ministry of Energy (SENER) of the Mexico Government. These data sets have substantial value as they can be leveraged to make better forecasts or detect anomalies more effectively, which in turn translates into a valuable strategy to aid forecast-based decision making. To expose the application of this tool, the definition, training, evaluation and visualization of the results of a deep learning model for time series was carried out using the Python programming language.