Early prediction of financial failure: An application in BIST

Karataş, B. (2024). Early prediction of financial failure: an application in BIST (Unpublished doctoral thesis). Sakarya University. The aim of this study is to develop models that can predict the financial success/failure of firms operating in the manufacturing sector in Borsa Istanbul (BIST). One year in advance prior to the onset of financial success/failure that is expected to persist for at least two consecutive years by utilizing statistical forecasting methods such as logistic regression analysis (LRA) and artificial intelligence-based models such as Fuzzy Logic (ANFIS). The scope of the research focuses on forecasting models developed to predict the financial success and failure of companies operating in the manufacturing sector in Borsa Istanbul (BIST). The research universe consists of companies operating in the BIST manufacturing sector in the period 2005-2019. The data was obtained from the 12-month (annual) financial statements of companies operating in the manufacturing sector in BIST. There are 177 companies in the manufacturing sector in BIST. Data was obtained from the 12-month (annual) financial statements of a total of 206 companies operating in BIST in the 2005-2019 period, 103 of which met the failure criterion and 103 of which met the success criterion. In the study, the financial failure criterion was taken as the basis for the companies' declaration of "Net Loss" for at least two or more consecutive years. There are 58 independent variables in the research. SPSS 22, Matlab R2019a, Jamovi 2.3.28 and Excel programs were used. The research consists of three parts. In the first part; the concept of financial failure, its reasons, its effects on businesses, prevention methods will be discussed. In the second part; the importance of early prediction of financial failure, models used in the prediction of financial failure and literature will be discussed. In the third part; an application on the BIST manufacturing sector on the prediction of financial failure will be found. The method of the research; the analysis of the data in the research was carried out with the logistic regression analysis method in the SPSS 22 program, and various statistical analyzes such as significance, explanatory power and fit of the model were included. The four independent variables that the prediction model presented with the logistic regression method found to have high explanatory power were also used as independent variables of the ANFIS method. The Matbab R2019a program was used for the ANFIS model. The ANFIS model was trained with the training data set, and predictions were made by showing other data sets that the network had never seen. The differences of the research from other studies are as follows; i) Working with a large mass of data over a long period (15 years) covering the years 2005-2019, ii) Developing a single model covering all sub-sectors rather than developing models separately for each sub-sector, iii) In addition to financial ratios, non-financial independent variables such as the free float rate of the company, the operating period of the company, and the opinion of the audit report should also be used, iv) Establishment of a hybrid model by operating the useful independent variables obtained from statistical methods with Fuzzy Logic-ANFIS, one of the artificial intelligence estimation methods, v) The first failure year of consecutive failures is considered as the financial failure period (t). Forecasts are calculated using the data of the year (t -1) preceding the base year of failure of the firms considered as financial failures. In the research, models presented with the Logistic regression and ANFIS method were compared. As a result, the study revealed that the ANFIS model produced better prediction results than the LRA, the most important independent variable used in the prediction of financial success/failure was Operating Profit or Loss (net)/short term liabilities. Keywords: Financial Failure, Financial Failure Prediction, Financial Failure Forecasting, Fuzzy Logic- ANFIS, Logistic Regression Analysis