Statement of the Problem
The obnoxious state of the Nigerian manufacturing sector has created a dire need for accurate bankruptcy prediction models about the overall outlook of companies. This is precipitated on the overbearing consequences of corporate bankruptcy on key stakeholders. Prior studies have mainly focused on the banking sector, using traditional statistical models, such as discriminant and ratio analysis (Nwidobie, 2017; Egbunike & Ibeanuka, 2015; Ezejiofor, Nzewi, & Okoye, 2014; Pam, 2013; Ebiringa, 2011; Usman, 2005), while few have addressed the manufacturing sector (Hur-Yagba, & Okeji, Ayuba, 2015; Ani & Ugwunta, 2012). Other studies have also demonstrated the practicality of logistic regression (Egbunike & Ezeabasaili, 2013). Despite the success of traditional statistical models they often violate certain assumptions, such as linearity, normality, multicollinearity, among others (Hua, Wang, Xu, Zhang, & Liang, 2007; Dimitras, Zanakis, & Zopounidis, 1996; Back, Laitinen, Sere, & van Wezel, 1996). They are often inadequate in identifying and estimating key parameters which limit their application in the real world (Hawley, Johnson, & Raina, 1990; Zhu & Rohwer, 1996). Secondly, the issue of time dimension limits the practicality of using previously developed models in present periods (Alaka et al., 2018). Bankruptcy prediction is a high-dimensional classification problem and most data distribution is non-Gaussian and exceptions are common (Zavgren, 1983). 7 The high-dimensional properties of data needed in model development also affect the classification accuracies of traditional statistical models (Zhang & Wu, 2011). Recent developments in artificial intelligence has widened its application to bankruptcy prediction problems, with the Neural Networks (NNs) being among the first (Alaka et al., 2018; Atiya, 2001; Wilson & Sharda, 1994, Serrano-Cinca, 1993; Coats & Fant, 1993; Udo, 1993). Studies have addressed the issue of bankruptcy among firms quoted on the Nigeria Stock Exchange using four widely acknowledged methods: discriminant analysis (Babatunde, Akeju, & Malomo, 2017; Ani & Ugwunta, 2012), logistic regression, probit regression (Adeyeye & Migiro, 2015) and neural networks (Yahaya, Nasiru, & Ebgejiogu, 2017; Farinde, 2013; Eriki & Udegbunam, 2013). Studies have confirmed the superiority of NNs to discriminant and logistic approaches (Eriki & Udegbunam, 2013; Farinde, 2013), with prior studies in Nigeria, focused on banks (Yahaya, Nasiru, & Ebgejiogu, 2017; Farinde, 2013), interest rate on loan investment (Enyindah & Onwuachu 2016), stock market (Eriki & Udegbunam, 2013), and insurance companies (Ibiwoye, Ajibola, & Sogunro, 2012). 8 NNs possess certain limitations, such as; the difficulty in building models as a result of many parameters to be set by heuristics. Secondly, is the danger of overfitting, and its lack of explanation ability, i.e., the ‘black box’ problem, as users do not also easily comprehend the final rules which the models acquire (Shin & Lee, 2002). However, an overall better performance model can only be achieved from an informed integration of tools to form a hybrid model (Alaka et al., 2018). Studies have shown that hybrid models have higher classification accuracies (Alaka et al., 2018; Bartual, Garcia, Guijarro, & Moya, 2013; Chen, Ribeiro, Vieira, Duarte, & Neves, 2011). The GA has been proved effective in developing hybrid models (Sai, Zhong, & Qu, 2007). A recent survey identified GA as one of the present data mining techniques that contribute to business decision making (Lin, Ke, & Tsai, 2017) and can provide new insights into bankruptcy prediction (McKee & Lensberg, 2002). Studies have underinvestigated the application of AI to the subject of bankruptcy prediction. In Nigeria application is limited to neural networks using feed forward and back propagation. The obvious lack of empiricism on the subject in developing countries stemmed the researcher’s interest on the subject. 9 Secondly, studies have questioned the reliability of models developed with only financial ratios, since there is doubt about the validity and reliability of the accounting information used for the ratios (Agarwal & Taffler, 2008). In addition, the relevance of particular ratios changes due to changes in the environment (Tsai, 2009). It may be worthwhile increasing the variety of explanatory variables to include corporate governance variables in developing prediction models (Ani & Ugwunta, 2012). Corporate governance structures are one of the prime causes of bankruptcy (Daily & Dalton, 1994; Gales & Kesner, 1994; Gilson, 1990; Hambrick & D’Aveni, 1988, 1992). Therefore the addition of corporate governance variables can improve the predictive power of bankruptcy models (Platt & Platt, 2012; Lajili, & Zéghal, 2010; Chang, 2009; Fich & Slezak, 2008; Donoher, 2004). However, the inclusion of corporate governance variables in GA selection and optimization process has been underinvestigated. According to Brédart (2014b) studies should be directed to this under-investigated aspect of corporate bankruptcy. Thirdly, in developing hybrid models GA has widely been applied in addition with other AI techniques (Min, Lee, & Han, 2006). This includes fuzzy logic and neural networks (Georgescu, 2017; Chou, Hsieh, & Qiu, 2017; Jeong, Min, & Kim, 2012; Esseghir, 2006); fuzzy Case Based Reasoning (CBR) method and Genetic Algorithms (Li & Ho, 2009); genetic-based support vector machines (GA-SVM) (Wu, Tzeng, Goo, & Fang, 2007; Min, Lee, & Han, 10 2006); Linear Genetic Programs (LGPs) (Mukkamala, Tilve, Sung, Ribeiro, & Vieira, 2006). Few studies have dealt with the integration of GA and Boosting Ensemble, such as the Gradient Decision Trees. One notable study is that of Sun and Hui (2006), applied decision tree and genetic algorithms for financial ratios' dynamic selection and financial distress prediction. Fourthly, most models rely on profitability ratios from financial statements which are prepared on an accrual basis. Therefore, they are deemed to be prone to aggressive accounting. However, in contrast ratios based on cash flow information is deemed to be more immune to manipulations (Welc, 2017). The study therefore also placed emphasis on cash flow ratios classified.
ABSTRACT
This study investigates foreign aid and its impact on socio economic development in Nigeria’s fourth...
Abstract
The study investigated on the effect of sex education on adolescents’ sexual behaviour among senior seco...
Abstract
This project is designed to give an insight on the influence of sales promotion on consumers loyalt...
ABSTRACT
This research examined Effect of Training and Development on Employee Performance with special reference to&nbs...
ABSTRACT
Business today is conducted in an age of highly complex technology imposing on banking officia...
ABSTRACT
Amaranthus viridis (Amaranthaceae), an annual herb, used in traditional medicine as diuretic and emollient, in poultices on infl...
Education is an important aspect of economic development. The social, economic and political needs of t...
ABSTRACT
The main thrust of this study is to examine the application of management accounting and its application to org...
ABSTRACT
This research work is aimed at investigating the influence of PTA on the...
ABSTRACT
The study examined the insecurity situation in Nigeria and its implications for business investment, opera...