Conditional Autoregressive Hilbertian processes

When considering the problem of forecasting a continuous-time stochastic process over an entire time-interval in terms of its recent past, the notion of Autoregressive Hilbert space processes (arh) arises. This model can be seen as a generalization of the classical autoregressive processes to Hilbert space valued random variables. Its estimation presents several challenges that were addressed by many authors in recent years. In this paper, we propose an extension based on this model by introducing a conditioning process on the arh. In this way, we are aiming a double objective. First, the intrinsic linearity of arh is overwhelm. Second, we allow the introduction of exogenous covariates on this functionvalued time series model. We begin defining a new kind of processes that we call Conditional arh. We then propose estimators for the infinite dimensional parameters associated to such processes. Using two classes of predictors defined within the arh framework, we extend these to our case. Consistency results are provided as well as a real data application related to electricity load forecasting.