Financial distress prediction based on serial combination of multiple classifiers

Abstract

Financial distress is the most synthetic form of business crisis and financial distress prediction (FDP) has been a widely and continually studied topic in the field of corporate finance. Recently, the advantage of FDP based on multiple classifiers’ combination began to be emphasized. This paper attempts to put forward a FDP method based on serial combination of multiple classifiers, which tries to make use of class-wise expertise of diverse base classifiers in serial combination system. Framework of serial combination system for FDP, selection mechanism of base classifiers and algorithm of FDP based on serial combination are discussed in detail. With financial condition dividing into two categories, empirical experiment indicated that FDP method based on serial combination of multiple classifiers performs at least as well as the best base classifier in average accuracy and stability, but it did not show much advantage in information complementation from base classifiers and was easy to be dominated by the first base classifier in serial combination system. This may be attributed to the number of target categories and serial combination method was inferred to be more suitable for FDP with multiple categories, which leaves to be further studied.