Finding Intrinsic Components in Economic Time Series

In many fields including economics, collection of time series such as stocks or energy prices are governed by a similar non-linear dynamical process. These time series are often measured hourly, thus, each day can be viewed as a high-dimensional data point. In this work, we apply a spectral method, which based on anisotropic diffusion kernels to model high dimensional electricity price data in the New York area. The purpose is to understand the major factors that affect the electricity prices and using those factors for this market`s needs such as prediction of the next price since this is a competitive market.

Today, predictions are made by using factors that are extracted by well-known linear, and sometimes simple non-linear methods. Using these factors cannot guarantee, even with cleaver prediction methods, that the found factors are indeed the most informative ones and that they provide all the information they have the potential to.

We demonstrate the proposed method on hourly marginal cost data that was collected from several zones. We show that even though the observed output spaces differ by local spatial influences and noise, the common global parameters that drive the underlying process can be extracted and separated from the unique zonal factors and the residual data. The major common factors that are extracted can be used to predict the next price since they are what mainly motivate the data.