The Investigation of Possibility of the Use of Genetic Algorithm in Predicting
Companies' Bankruptcy
(Experienced Evidence in Iran)
Mansour Garkaz
Accounting dept; Islamic Azad University;
Aliabad Katoul Branch, Golestan Province,
IRAN
Ahmad Abdollahi
Accounting dept; Golestan Non- Governmental Non for
Profit Institute of Higher Education; Golestan Province,
IRAN
Abstract-The purpose of this paper is to predict the
bankruptcy using genetic algorithm. So doing, having
read the review of literature, the researchers found a
complete list of financial proportions that showed high
capabilities the predicting bankruptcy. These
proportions include the ratio of operational income to
sale, ratio of total debts of total assets; current assets to
current debts; sale to current assets and interest cost to
grass profit. Then the possibility of the use of the genetic
algorithm in predicting the bankruptcy of the accepted
companies in Tehran Stock Exchange was investigated.
The independent t-test showed that there was a
meaningful difference between the average of these
ratios of bankrupted group with that of non-bankrupt
one. The results of the statistical tests indicate that the
genetic algorithm can be used of predict bankruptcy in
Iran.
Key words: predicting bankruptcy; genetic algorithm,
Tehran Stock Exchange; type 1 and 2 Error.
I. I. INTRODUCTION
The economic crisis of the 1930s necessarily led
financial studies to final saving aspects for the companies
and the issues such as liquidity, bankruptcy, liquidation and
refounding the companies [7]. The world economy has
become aware of the risk within the capital structure of the
companies especially after the bankruptcy of some huge
companies such as WorldCom and Enron. If we can cull
information pertaining to the possibility of bankruptcy after
the real occurrence, we can decrease an even prevent its
economic and social consequences. Thus the correct
predicting of bankruptcy in the financial world is of vital
importance [2]. By the way models to predict bankruptcy,
we means these models that are able to predict the
bankruptcy and decrease investing risk using financial
information released by accounting [8]. Bankruptcy
predicting models can be divided into two main groups. The
first group is those models that predict bankruptcy using
information of the market and analyzing them and the 2
nd
group refers to these ones that predict bankruptcy using
accounting information (financial ratios). The 2
nd
group
models can be divided using modeling method into statistical
models, models based on artificial intelligence and
theoretical models. Table 1 shows different predicting
paradigms. In all the models in use, the suitable model
should be capable enough to classify the companies correctly
[4]. The recent studies pertaining to the genetic algorithm
rarifies the fact that this algorithm due to nonlinear
nonparametric qualities and comparative learning is a
powerful tool to identify and classify the models [10]. That
is why this kind of algorithm has been used in this study.
In a research entitled "The application genetic algorithm
in the analysis of bankruptcy danger", Varetto f. (1998) has
made use of the genetic algorithm in predicting bankruptcy.
The studies included 500 companies consisting of 236
bankrupted and 264 not bankrupted ones. The findings
indicated the 93% predicting accuracy one year before
bankruptcy and 91.6% accuracy two years before bankruptcy
[1].
Shin and Lee (2002) also presented a model beside on
genetic algorithm that showed how this algorithm could be
used in modeling predicting the bankruptcy. They
investigated 528 producing comprised of 264
bankrupted and 265 not bankrupted from 1995 to
1997. Their results showed that the designed
model could predict bankruptcy one year in
advance its occurrence with the accuracy of 80%
[6]. Min and Jong (2008) suggested a new
classification based on the genetic algorithm to
predict bankruptcy. The proposed method was also
flexible and was able to be applied in other fields such as
predicting the purchase of the products or risk management
of the project. The financial ratios of 2542 small and middle
audited producing companies and the same number not
bankrupted one were used as the data of this research [5].
Considering all the research done, it can be understood that
although the statistical models could predict the bankruptcy
well, some limiting assumptions such as linearity. Normality
and independent relation among the variables could affect
the efficiency of these models. Therefore, other methods
have been introduced to overcome some or even all of these
limitations to improve the predicting performance
The research hypothesis is as follows:
The genetic algorithm is able to predict bankruptcy of
the companies in Iran Stock Exchange.
282
2010 International Conference on Business and Economics
Research
vol.1 (2011) © (2011) IACSIT Press, Kuala Lumpur, Malaysia