Using of Non-financial Data in Predictive Models
Using of Non-financial Data in Predictive Models

Author(s): Zuzana Kubaščíková, Miloš Tumpach, Zuzana Juhászová
Subject(s): Business Economy / Management, Methodology and research technology, Accounting - Business Administration
Published by: Masarykova univerzita nakladatelství
Keywords: financial statements; financial analysis; neural network; deep learning; data mining;
Summary/Abstract: Forecasting the company's future economic situation arose in the early 20th century. First of all, a multidimensional discriminatory analysis was used to construct prediction models, later replaced by logistic regression. The new challenge in predicting financial development is neural networks representing a more reliable financial forecast compared to mathematical and statistical methods. The neural network, by mimicking the capabilities of human brain neurons, is capable of modeling the course of dependencies between individual indicators and results. A disadvantage of the original prediction models is also the low range of empirical accounting data and the fact that they are focused only on financial data. The introduction of the financial statements registers led to the possibility of free access to full data from the financial statements, which opens the door to new possibilities in scientific research. For the purpose of this paper an annual report of 20 selected companies were tested. The aim of this paper is to accept or reject the claims that non-financial „narrative” data could be also used for the assessment of the financial position and financial performance of the companies. The results of our sentiment analysis supported the hypothesis that financially distressed companies use a different tone of language in their annual reports compared to financially stable companies. These findings confirmed the relationship between the tone which managers use in constructing annual report narratives, and the financial performance of the company. Therefore, it is advisable to incorporate non-financial data into the forecasting models.

• Details
• Contents