DOI: 10.5593/sgem2017/21/S07.012


C. Doru, E. Clipici
Wednesday 13 September 2017 by Libadmin2017

References: 17th International Multidisciplinary Scientific GeoConference SGEM 2017, www.sgem.org, SGEM2017 Conference Proceedings, ISBN 978-619-7408-01-0 / ISSN 1314-2704, 29 June - 5 July, 2017, Vol. 17, Issue 21, 85-94 pp, DOI: 10.5593/sgem2017/21/S07.012


The aim of this paper is to propose an improved algorithm for estimating the risk of bankruptcy of the Bulgarian insurance companies. The algorithm is developed based on the multi-layers neural arhitecture with input dates represented by values of indicators from the financial statements of Bulgarian insurances companies. The memory of the neural network will provide informations about the risk of bankruptcy. Development of the neural model (algorithm R2b2n- risk bankruptcy of Bulgarian insurance companies with neural network) to estimate the risk of bankruptcy for insurance societies is done according to the following steps:
1. forming neuronal structure by an neural multi-layers architecture with three layers of neurons - an input layer neurons to take the training data, one hidden layer for information processing and an output layer neurons of the neural network to formulate response of the neural evaluations;
2. created database with the input data represented by the financial-economics indicators from the financial situations of the Bulgarian insurance companies. Here we considers 31 bulgarian insurance companies and 7 fundamentals parameters for financial indicators; economic indicators values were taken from financial situations published on the official website of the the Financial Supervision Commission (FSC) - an institution that is independent from the executive authority and reports its activity to the National Assembly of the Republic of Bulgaria and is a specialized government body for regulation and control over different segments of the financial system, capital market, insurance market, health insurance market, pension insurance market.
3. establish the neural network training data and setting the testing data of the R2b2n algorithm based on indicator values for each firms for many years (2013-2016).
4. determining the weighting map for estimating the risk of bankruptcy based on the encoded values of the neural network. The neural map for estimating the risk of bankruptcy is determined based on neuronal synaptic weights and is calculated on the basis of neural threshold and average values, respectively, standard deviation of the responses of neural network. The neuronal map for estimating contains four risk areas: a minimal risk region, an area of low risk of bankruptcy, an area of high risk of bankruptcy and the risk of imminent bankruptcy. This map is related to the results of the Altman model (1968) which sets two limit areas (one without risk and another of bankruptcy) and an area of uncertainty (high risk and low risk).
In the finally of the paper is performed the testing capacity for estimate the risk of bankruptcy for all Bulgarian insurance companies by applying neural function determined by R2b2n algorithm, and establishing a synthetic report on risk areas associated with each firm, and highlighting the evolution trends of neural function score estimation in the next year for Bulgarian insurance companies. The estimation algorithm developed is distinguished by the property to generalize; the neuronal map that computing the risk of bankruptcy may be determined by applying the algorithm R2b2n both for the Bulgarian insurance-reinsurance companies and for to the global insurance-reinsurance companies.

Keywords: Artificial Neural Network, Backpropagation, Risk of Bankruptcy, Insurance and Reinsurance

Home | Contact | Site Map | Site statistics | Visitors : 0 / 353063

Follow site activity en  Follow site activity INFORMATICS  Follow site activity Papers SGEM2017   ?

CrossRef Member    Indexed in ISI Web Of Knowledge   Indexed in ISI Web Of Knowledge

© Copyright 2001 International Multidisciplinary Scientific GeoConference & EXPO SGEM. All Rights Reserved.

Creative Commons License