Time series forecasting is critical for several business activities, and the extraction of features such as trend and seasonality is crucial for forecasting models accuracy.

“Trend and seasonality features extraction with pre-trained CNN and recurrence plot” (by Fernanda Strozzi and Rossella Pozzi, published in International Journal of Production Research) contributes to facilitating time series feature extraction by proposing a method that applies deep learning techniques to business time series images.

The GoogLeNet, a pre-trained Convolutional Neural Network that allows transfer learning and has achieved high recognition rates in image classification tasks, is applied to time series images obtained with a Recurrence Plot, an imaging method that depicts the recurrence of the state space system using coloured points and lines in 2D images. This application of deep learning techniques to business time series imaging provides interesting results and offers opportunity for further developments.