Deep learning for time series

The main purpose of this technical report is only to introduce to the use of the "pandas" open source, BSD-licensed library1. and of the TensorFlow framework2 fot forecasting time series. The use case chosen is of great relevance, the application is a mere experiment that a posteriori can find a confirmation: data is used for SARS Cov-2 spread forecasting too, it is fair to reiterate that these are only a kind of study notes. This is not an epidemiological study, moreover forecasts are strongly based on the numbero suspected cases, that is most probably higher than reported, of course on the behaviour of citizens, and on other unpredicatble factors. Moreover a recent study3 supposes that most infective people are asympthomatic.