Open Forecast is a tool (container) to automatically perform, based on time series, one or more products, using machine learning algorithms.
Our forecast engine offers an approach to the world of artificial intelligence and deep learning in a simple and concrete way, where the only user interaction is at the stage of entering data and characteristics of the desired forecast.
The system developed, through an exploratory analysis, detects anomalies in the series, such as missing values or temporary gaps, for the user who defines the treatment of those values.
Subsequently, it is responsible for finding the best model that fits the behavior of the series, evaluating mathematical algorithms, of different nature, taking into account the use of non-linear functions, which capture complex elements in the dynamics of a time series.
As a result, a forecast and the prediction intervals of the series are provided, complementary to the performance metrics of the model used and a descriptive statistic, a general and weekly level, which gives an overview of the current status of each product.