A framework for seismic risk policy design and the assessment of avalanche-like event prediction in knitted fabric.

Knitted fabric exhibits avalanche-like events when deformed, and as in most analogous seismic models, the occurrence of those time-intermittent and scale-invariant events cannot be accurately predicted, due to their peculiar statistics. We apply time-series prediction with Neural Networks on the mechanical response of the fabric in order to forecast future quantities. Furthermore, we discuss the assessment of such predictors, not only with standard metrics like the learning accuracy but also with more advanced ones taking into account the increase of risk with the amplitude. We present a framework allowing for the design of decision-making policies, as both a predictor evaluator and a provider of a clear strategy when dealing with risks, with KnitCity: a model seismically active city.