Text-based machine learning improves distress modelling
Analysis - January 2019 - No. 3
Machine learning methods make data modelling more flexible and enable the use of unstructured data, which conventional statistical methods do not permit to the same extent. If such methods are used to include auditors' reports and managements' statements from firms' annual reports when calculating their probability of distress, the accuracy of the calculation is increased considerably. Especially the auditors' reports contribute useful information.