Forecasting the Financial Bankruptcy of Iranian Listed Companies Using a Hybrid DEA–PCA Approach

Abstract

The topic of predicting company bankruptcy has attracted significant interest among financial researchers and experts. Due to the considerable impact of financial distress on companies' stakeholders, the development of accurate methods and models for forecasting bankruptcy and financial failure remains a key area of financial research. Investors consistently expect their capital to be secure and to receive returns that reflect the risks undertaken. Furthermore, the capacity to predict financial crises in companies in a timely manner in order to prevent capital loss is of critical importance. To address this need, researchers have conducted extensive studies employing various models and methods to evaluate corporate financial performance and forecast bankruptcy. However, it is essential to note that no single method is sufficient on its own; the best outcomes are achieved by combining multiple approaches with expert professional judgement. One technique that has gained increased attention in recent years for facilitating financial decision-making processes is Data Envelopment Analysis (DEA), which has produced acceptable predictive results. In this study, 52 manufacturing companies listed on the Tehran Stock Exchange were selected from three sectors: food and pharmaceuticals, metals, automotive and machinery, and chemicals and petrochemicals. Specifically, the first group included 21 companies (10 bankrupt and 11 healthy); the second group included 18 companies (10 bankrupt and eight healthy); and the third group included 13 companies (7 bankrupt and six healthy). The primary objective of this research is to evaluate the DEA model's ability to predict bankruptcy, i.e., to classify companies according to their financial distress status. To improve the performance of the DEA model, Principal Component Analysis (PCA) was used to reduce the dimensionality of its input variables.