LOUKERIS N.(1), MATSATSINIS N. (2)
Department of Production Engineering and Management,
Technical University of Crete,
Akrotiri Campus, 73100 Chania, Crete, Greece
(1)PhD student, Essex University, (2) Associate Professor, Technical Univ., of Crete
The amount of accounting data quite often is extended and consist a significant disadvantage for the professional analysts or the financial experts that usually evaluate businesses. Informatics, computational neurosciences and the science of finance created original tools capable of dealing with large economic data in an effective manner. An introduction to the most recent methodologies that can be used in the financial analysis of businesses includes: a)Data Mining, b) Neural Networks, c) Evolutionary algorithms, d) Theory of the Rough Sets, e) Theory of the Fuzzy Sets, f) Multicriteria Decision Analysis g) Classical Financial Analysis. The objective of this research is to evaluate the financial status of companies, to investigate the possibility of exante preference, aiming to estimate possible future earnings or losses from investments. Artificial Intelligence methods either classify companies in known groups, with different level of risk, or cluster in homogenous sets with similar attributes. Classification with data mining techniques is compared to the initial classification by bank executives in the loan department of a Greek commercial bank, in order to verify its steadiness, through supervised training, given the initial estimations. In cases of clustering the data, initial classifications are not taken in to account, functioning non supervised training on the financial indices and on the last stage creating clusters that describe companies with identical characteristics. The processes of classification, clustering and discrimination include techniques of Artificial Intelligence-Data Mining, Hybrid methods such as Neurofuzzy Logic and Neural-genetic Networks, and finally Neural Networks, in a sample from the loan portfolio of the same bank, taking into consideration 1411 companies from different areas of activity with data from the period 1994-1997.
To read this full paper please click the "Download Document" button below: