Financial Applications of Self-Organizing Maps

Applications of neural networks to finance and investments can be found in several books and articles [5]. The great majority of these applications use supervised neural network models for forecasting market trends, creating trading models, portfolio or risk management. So far few applications of unsupervised neural networks in finance are documented in the literature. Nevertheless unsupervised neural networks haven proven to be very successful in other fields [7]. A vast number of applications can be found in T. Kohonen’s Self-Organizing Maps [8]. This article provides an introduction to the use of self-organizing maps in finance, in particular it discusses how self-organizing maps can be used for data mining and discovery of patterns in large data sets. The illustrations provided include the selection of mutual fund investment managers, mapping of investment opportunities in emerging markets, and analysis of country risks. This article is based on a comprehensive review of financial applications of self-organizing maps summarized in a book tha t will be published in 1998 and is edited by the author in collaboration with T. Kohonen [6]. Exploratory data analysis and data mining Exploratory data analysis and data mining [1, 2, 3 ] are used for knowledge discovery in large data bases (KDD). The emphasis in exploratory data analysis is on the whole interactive process of knowledge discovery, i.e. the discovery of novel patterns or structures in the data. There is confusion about the exact meaning of the terms 'data mining' and 'knowledge discovery'. At the first international conference on knowledge discovery in Montreal in 1995 it was proposed that the term knowledge discovery be employed to describe the whole process of extraction of knowledge from data. In this context knowledge means relationships and patterns between data elements. It was further proposed that the term d a t a mining should be used exclusively for the discovery stage of the process.