CV: Curriculum Vitae from Carlos Serrano Cinca, professor in Accounting and Finance at the University of Zaragoza (Spain)

Serrano Cinca, C. (1998): "Self-organizing Maps for Initial Data Analysis: Let Financial Data Speak for Themselves", in Visual Intelligence in Finance using Self-organizing Maps, julio 1998, Ed Guido Deboeck & Teuvo Kohonen, Springer Verlag


In this Chapter we consider the possibility of using self-organising maps (SOM) as a complementary technique for IDA. As we have already noted from the detailed descriptions contained in the first chapters of this book, SOM takes an initial data set to which it applies a process known as self-organisation. It can be extremely useful as an IDA technique with financial and economic information, allowing the data to speak for themselves. When financial information on a group of companies is introduced, these companies will be self-organised in such a way that those with similar financial characteristics will be located close to one another on the map. Additionally, SOM allows us to study the evolution of a company over time by introducing information coming from different accounting periods, to place it in relation to its competitors and to prepare sectoral maps, as well as offering us many other possibilities which we shall go on to consider.
In this Chapter we shall apply SOM to company solvency, to bond rating, to the strategy followed by the company in relation to the sector in which it operates on the basis of its published accounting information, and to the comparison of the financial and economic indicators of various countries. Naturally, allowing the data to speak for themselves does not mean that this analysis is sufficient. All IDA is an important step within any piece of empirical research, but it is no more than a first step and it must be completed with other analyses. In this Chapter we shall study the integration of SOM into a decision support system (DSS) that is useful for predicting the probability of company bankruptcy. Employing SOM does not imply that the use of other well-known techniques is renounced; rather, it is very productive to complement it with other tools. Indeed, we will complete and compare SOM with multivariate statistical models such as Linear Discriminant Analysis (LDA), as well as with neural models such as the Multilayer Perceptron (MLP).


Neural Networks; Multilayer Perceptron; Self-organizing Maps; Initial Data Analysis

Download (Amazon) (draft in word)