Decision trees to identify companies’ distress: the AI at work
Frame of the research: The main subject of investigation is represented by the valuation of a company’s distress, adopting decision trees, a well-established artificial intelligence methodology to automatically identify a combination of attributes to explain two target variables of interest, that is the zone of discrimination and cut off. The proposed methodology allows for the representation of decision processes according to paths on the tree’s branches, or through a set of easily browsable if-then rules.
Objectives: The study aims to examine whether and how artificial intelligence (AI) may facilitate the joint comprehension of corporate distress and corporate legality. The main subjects of investigation are both represented by the valuation of the company’s distress, as well as by the legality rating (LR), which is a measure of the company’s degree of legality. The combination of a new set of variables, allows to understand - within a given range of accuracy - the company’s financial health, and conversely, the company’s distress, regardless of the Altman Z-score.
Methodology: The dataset is composed of companies in possession of legality ratings. Two experimental settings, which make use of decision trees, allow us in this study to automatically identify the unique combination of variables from the dataset that explains two target variables - ‘zone of discrimination’ and ‘cut off’ - from the standpoint of a unique perspective; one that is not considered by the Altman Z-score.
Findings: AI allows for the identification of a new ‘basket’ of variables, one different from those employed by the Altman Z–score. These variables may be used to determine a company’s level of distress. The experiments test the ‘ability’ of the algorithm to identify a combination of variables to predict the target value. It is thereby possible to analyse in which way these variables operate alongside one another to produce with accuracy the correct identification of the target variable. In light of this scenario, the contribution of the study is the identification of two algorithms able to determine two settings of if-then rules that produce the same outcomes obtainable through the application of the Altman Z-score model, without directly using the model itself.
Research limits: The methodology described above was required to determine a plausible interval for the variables identified by the decision trees. The current development of the research, however, reveals that the methodology still needs to be adapted in order to determine the plausible intervals for the variables identified by the decision trees. In fact, the dimensionality of the dataset could benefit from resampling the variables for the proposed methodology, which suffers from certain degrees of skew.
Practical implications: The AI methodology can process large amounts of records within a given dataset, thereby allowing for the testing of the effectiveness of LR in the assessment of creditworthiness.
Originality of the study: The recognition and composition of the new variables can be interpreted as a tool to strengthen the comprehension of the company’s distress.
ALTMAN E.I. (1968), “Financial Ratios. Discriminant Analysis and the Prediction of Corporate Bankruptcy”, The Journal of Finance, n. 23, pp. 589-609.
ALTMAN E.I. (1983), Corporate Financial Distress, Wiley Interscience, New York
ALTMAN E.I. (2000), Predicting Financial Distress of Companies: Revisiting The Z-Score and Zeta, Handbook of Research Methods and Applications in Empirical Finance, 5.
ALTMAN E.I., HARTZELL J., PECK M. (1995), Emerging Markets Corporate Bonds: A Scoring System, Salomon Brothers Inc. New York.
ALTMAN E.I., HOTCHKISS E. (2006), Corporate Financial Distress and Bankruptcy, 3rd edition, Hoboken, NJ, J. Wiley & Sons.
ALTMAN E.I. (2013), “Predicting financial distress of companies: revisiting the Z-Score and ZETA® models”, in Handbook of Research Methods and Applications in Empirical Finance, Cheltenham, UK: Edward Elgar Publishing.
ANDERSON D.R., SWEENEY D.J., WILLIAMS T.A., CAMM J.D., COCHRAN J.J. (2015), An introduction to management science: quantitative appROAches to decision making, Cengage Learn.
BARILE S., PICIOCCHI P., BASSANO C., SPOHRER J.C., PIETRONUDO M.C. (2019), “Re-defining the role of artificial intelligence in wiser service systems”, Kacprzyk J. (Ed.), Advances in Intelligent Systems and Computing (AISC), Vol. 787, Springer International Publishing AG, part of Springer Nature, Switzerland, pp. 159-170.
D’AVANZO E., D’ANTÒ V., MICHELOTTI A., MARTINA R., ADINOLFI P., PANGO MADARIAGA A.C., ZANOLI R (2018), “A Collaborative Web Service Exploiting Collective Rules and Evidence Integration to Support Sustainable Orthodontic Decisions”, Journal of Cleaner Production, n. 176, pp. 813-826.
DAMODARAN A. (2002), Investment valuation: Tools and Techniques for Determining the Value of Any Asset, John Wiley and Sons, Inc.
MASÍAS V.H., KRAUSE M., VALDE S.N., PEREZ J.C., LAENGLE S. (2015), “Using decision trees to characterize verbal communication during change and stuck episodes in the therapeutic process”, Front. Psychol., n. 6, p. 379.
MITCHELL T.M. (1997), Machine Learning, 1 ed. McGraw-Hill, Inc., New York, NY, USA.
OHLSON J.A. (1980), “Financial ratios and the probabilistic prediction of bankruptcy”, Journal of Accounting Research, vol. 18, n. 1, pp. 109-131.
PRATT S.P., GRABOWSKI R.J. (2010), Cost of Capital - Fourth edition, John Wiley and Sons, Inc., New York.
QUINLAN R.J. (1993), C4.5: Programs for Machine Learning, Morgan Kaufmann Publishers Inc., San Francisco, CA, USA.
VULPIANI M. (2014), Special Cases of Business Valuation, McGraw-Hill, Milano.
WITTEN I.H., FRANK E. (2011), Data Mining: Practical Machine Learning Tools and Techniques, Morgan Kaufmann.