SMOOTH COMPONENT EXTRACTION FROM A SET OF FINANCIAL DATA MIXTURES

摘要

Independent Component Analysis (ICA) is considered in this paper, which is a Signal Processing method for expressing an observed set of random vectors as a linear combination of statistically independent components. There have been numerous successful implementations to ICA, each using a different interpretation of independence as the objective. The application of a recently developed sequential blind signal extraction algorithm is examined, which apart from the negentropy cost function has an additional constraint aiming to identify smooth independent components in the data set. The signals examined originate from the financial markets and are a portfolio of technology stocks from the NASDAQ stock market. The resulting independent components (Ics) are examined and contrasted to those obtained through a traditional implementation of ICA. Adopting several viewpoints, possible advantages of this novel approach are demonstrated over FastICA, when searching for the underlying sources that give rise to stock evolutions and their structure in time.