Lietuvos kredito unijų veiklos rizikingumo vertinimas

LTStraipsnyje apibūdinamos Lietuvos kredito unijų veiklos rizikingumo vertinimo galimybės taikant skirtingus daugiamatės analizės metodus (sprendimų medžių CART, CHAID ir išsamiojo CHAID bei dirbtinių neuroninių tinklų modelių), atliekamas jų lyginimas, pagrindžiamas patikimumas ir tikslingumas vertinant gautus rezultatus, taip pat tikrinamas modelių stabilumas laiko požiūriu, įvertinamas gautų rezultatų patikimumas ir pagrindžiamas šių modelių tinkamumas tiriant Lietuvos kredito unijų bankroto galimybes. [Iš leidinio]Reikšminiai žodžiai: Dirbtiniai neuroniniai tinklai; Kredito unijos; Rizikingumas; Sprendimų medis; Sprendimų medis dirbtiniai neuroniniai tinklai; Vertinimas; Artificial neural networks; Classification trees; Credit unions; Estimation; Riskiness.

ENThe main purpose of this paper is to develop models that would name risky credit unions, in other words, credit unions, which are most likely at risk of bankruptcy. The work consists of two main parts: the analysis of literature and the research, and its results and conclusions. When the survey of the literature was carried out, the authors made financial indicator sets which were used for classification of the credit unions into the risk groups. Bankruptcy cases in Lithuania were insufficient so the authors suggested two ways to measure credit unions’ riskiness. Based on good classification results of the surveyed researchers the authors have chosen decision trees and artificial neural network methods to solve a classification problem. Decision trees were formed using CART, CHAID and exhaustive CHAID analysis. With these methods applied, some research was carried out using distinct financial indicator sets and different credit union classification in risk groups. The performed research revealed that the most significant financial indicators classifying risky credit unions were net profit and share capital ratio, capital and asset ratio, income and capital ratio, also share capital and asset ratio, net profit and average asset ratio, loans and capital ratio. The best classification accuracy was achieved by using artificial neural networks. With different classification of credit union riskiness, the most important financial indicator was interest paid on deposits and average market interest rate ratio. The best classification accuracy was achieved by decision tree made by CART analysis. The authors believe that the results of the study could provide credit union members (or potential members) with useful guidelines regarding credit unions which they should avoid and which not. Moreover, this information could be useful for supervisors. [From the publication]