Ensemble Methods for Jump-Diffusion Models of Power Prices

We propose a machine learning-based methodology which makes use of ensemble methods with the aims (i) treating missing data in time series irregular observation times and detecting anomalies observed behavior; (ii) defining suitable models system dynamics. applied this to US wholesale electricity price that are characterized by data, high stochastic volatility, jumps pronounced spikes. For we provide repair approach based on missForest algorithm, an imputation algorithm is completely agnostic about distribution. To identify anomalies, i.e., turbulent movements power prices spikes observed, took into account no-gap reconstructed series, then detected anomalous regions using isolation forest anomaly detection method isolates instead profiling normal points as most common techniques. After removing additional gaps will be newly filled algorithm. In way, complete clean describing stable dynamics can obtained. The decoupling between motion allows us define jump-diffusion estimation procedure uses full information contained both