Financial distress prediction using the Q&A text of online interactive platforms

Abstract

The text-based question and answer (Q&A) information of online interactive platforms reflects the concerns of investors and the responses of companies, which can reduce information asymmetry and risk accumulation. This paper proposes a framework to predict the financial distress of companies and explores whether the features extracted from Q&A text can significantly improve the performance of financial distress prediction (FDP) models. In this framework, we extract interactive features, a question sentiment feature, an answer satisfaction feature and topic features. Experimental results show that interactive features and topic features can significantly improve the performance of FDP models compared with the model using only financial features. In addition, we apply the Shapley value method to identify the key factors in predicting the financial distress of companies, providing a basis for investors and managers to make decisions.