GOING CONCERN MODELS AND TECHNIQUES
The first method that was developed for the purposes of ascertaining the bankruptcy is the multivariate discriminate analysis which is also referred to as MDA, logit analysis, neural networks. These method were considered to be used for the primary methods for the model development. MDA goes in for the classification of the various firms into groups which is somewhat based upon the characteristics of each of the ratios and the factors of the firm. This is based upon the sample observations. The various coefficients have been calculated for each of the characteristic. There are many of the coefficients that is the product of the various ratios and their coefficients and this helps in giving us the discriminant score which allows the classification of the firm.
The logit analysis and the probit on the other hand considers the probability that the firms would either be classified as the going concern or the non-going concern. The main difference between these two is the fact that probit analysis requires the non-linear estimation whereas logit analysis does not. There are number of different method that help in finding this correlation.
There were many of the neural networks that helps in the analysis of the various different patterns and this helps in the development of the model which is somewhat capable of the process of decision making.
Bellovary, J., Giacomino, D. and Akers, M. (2007). A Review Of Going Concern Prediction Studies: 1976 To Present. [online] Epublications.marquette.edu.
The next method or the technique is based upon the explanatory power of the bankruptcy of the prediction models. The researchers never undertake the measurement of the information which has been observed by the auditor and the lenders of the client. This information is used by the auditor and the lenders but somehow goes unobserved by these researchers and that enters in the terms of the equation of bankruptcy and the equation of the going concern and this causes them to be correlated. There is a bivariate probit that identifies the direct effect of the various opinions of the going concern and this holds good even where are errors in the various correlated terms and this is somewhat worth of controlling for the observable factors that are possible for the following 3 reasons:
- It helps in the reduction of the noise of an error in terms of the estimator which is considered to be more efficient. This leads to an increased power of the various empirical tests.
- There is a reduction of the correlation between the 2 error terms. Such of the reductions would limit the reliance on the various assumption that allows the joint normal distribution. In this limit, there is an observed and a controlled for all the information that has bene observed and that has been used by the auditor. This helps in the identification of the causal effects on the opinion on the going concern on bankruptcy along with the equation of the bankruptcy. This goes in the absence of the imposing of the distributional assumptions in the terms of an error.
- The linear model could also be mis-specified. This is likely to be bias in the terms of the estimations.
There is an inducement of the results in the discontinuity for the purposes of the linear model and that would not be taken into account.
Gerakos, J., Hahn, P., Kovrijnykh2, A. and Zhou, F. (2015). The effect of going concern opinions: Prediction versus inducement∗. [online] home.uchicago.edu.
The third method is the least absolute shrinkage and the selection operator. In this, there is a stepwise regression which is applied in the work that is related with the past but there are many issues related with it.
The fourth method is the Support vector machine (SVM) which was developed by Boser during the year 1992 for the purposes of providing some better solutions than the other traditional classifiers. This involves the neural networks. This method (SVM) is one of the maximal margin classifiers which concerns the classification of the problem of the process of optimization.
The fifth method is the Class and the regression tree. This is a method which is used to describe the variable Y which has been distributed after the assignment of the forecast vector X. this method is able to segregate between the division rule wherein there is an identification of the data which is value and also it helps in achieving the ideal results.
Goo, Y., Chi, D. and Shen, Z. (2016). Improving the prediction of going concern of Taiwanese listed companies using a hybrid of LASSO with data mining techniques. SpringerPlus, 5(1).
Then there are the method of data mining which contains a large number of algorithms which have been derived from the osmosis of the statistics. The following are the various methods.
Kirkos, E. and Manolopoulos, Y. (2016). DATA MINING IN FINANCE AND ACCOUNTING: A REVIEW OF CURRENT RESEARCH TRENDS. [online] delab.csd.auth.gr.
- Genetic Algorithms Genetic Algorithms (GA): this is the method which is connected with the application of the ideas from the various natural evolution where the fittest individuals are able to survive. There are many of the rules that concerns with the encoding of the set of the strings each one of which consist of the bits. These strings forms the part of the population. This method allowed the strings that has the highest value of fitness and which helps in surviving and also proliferate the renewal of the population.
- Decision trees: this is the method which helps in prediction and involves the observations into the mutually exclusive subgroups. This method helps in searching for the various attributes for the best which helps in separation of the various individual classes. These subgroups are divided into the subgroups until and unless they are way too small and there is no statistical difference between these subsets. In case, the decision tree becomes too large then the method finally becomes pruned.
References:
Bellovary, J., Giacomino, D. and Akers, M. (2007). A Review Of Going Concern Prediction Studies: 1976 To Present. [online] Epublications.marquette.edu. Available at: http://epublications.marquette.edu/cgi/viewcontent.cgi?article=1002&context=account_fac [Accessed 10 Oct. 2016].
Gerakos, J., Hahn, P., Kovrijnykh2, A. and Zhou, F. (2015). The effect of going concern opinions: Prediction versus inducement∗. [online] home.uchicago.edu. Available at: http://home.uchicago.edu/~frankzhou8674/GoingConcerns20151114.pdf [Accessed 10 Oct. 2016].
Goo, Y., Chi, D. and Shen, Z. (2016). Improving the prediction of going concern of Taiwanese listed companies using a hybrid of LASSO with data mining techniques. SpringerPlus, 5(1).
Kirkos, E. and Manolopoulos, Y. (2016). DATA MINING IN FINANCE AND ACCOUNTING: A REVIEW OF CURRENT RESEARCH TRENDS. [online] delab.csd.auth.gr. Available at: http://delab.csd.auth.gr/papers/ICESA04km.pdf [Accessed 10 Oct. 2016].
