Data Envelopment Analysis may Obfuscate Corporate Financial Data: Using Support Vector Machine and Data Envelopment Analysis to Predict Corporate Failure for Nonmanufacturing Firms
Loading...
Date
2017
Authors
Yang, Xiaopeng
Dimitrov, Stanko
Journal Title
Journal ISSN
Volume Title
Publisher
Taylor & Francis
Abstract
Corporate failure prediction has drawn numerous scholars’ attention because of its usefulness in corporate risk management, as well as in regulating corporate operational status. Most research on this topic focuses on manufacturing companies and relies heavily on corporate assets. The asset size of manufacturing companies play a vital role in traditional research methods; Altman’s 𝑍 score model is one such traditional method. However, a limited number of researchers studied corporate failure prediction for nonmanufacturing companies as the operational status of such companies is not solely correlated to their assets. In this paper we use support vector machines (SVMs) and data envelopment analysis (DEA) to provide a new method for predicting corporate failure of nonmanufacturing firms. We show that using only DEA scores provides better predictions of corporate failure predictions than using the original, raw, data for the provided dataset. To determine the DEA scores, we first generate efficiency scores using a slack-based measure (SBM) DEA model, using the recent three years historical data of nonmanufacturing firms; then we used SVMs to classify bankrupt and non-bankrupt firms. We show that using DEA scores as the only inputs into SVMs predict corporate failure more accurately than using the entire raw data available.
Description
This is an Accepted Manuscript of an article published by Taylor & Francis in INFOR: Information Systems and Operational Research in 2017, available online: https://doi.org/10.1080/03155986.2017.1282290
Keywords
support vector machine (SVM), data envelopment analysis (DEA), corporate failure predictions, nonmanufacturing firms, data obfuscation