Deep learning based regime-switching models of energy commodity prices

Abstract We discuss a deep learning based approach to model the complex dynamics of commodity prices observed in real markets. A regime-switching is proposed describe time evolution market prices. In this model, base regime described by mean-reverting diffusion process and second driven predictions neural network trained on log-returns series. statistical technique, method simulated moments, estimate data. applied methodology energy price series with very different characteristics, namely US wholesale electricity, natural gas crude oil daily The obtained results show good agreement empirical particular, seems reproduce interesting way first four central moments distributions as well shape