Predicting Business Vulnerability Using Neural Networks: Evidence From Brazilian SMEs

Abstract

Goal: Analyze the performance of a business diagnosis prediction model. Design / Methodology / Approach: Business diagnostic interviews were carried out with 98 companies, using the IncubE methodology, to assess the level of maturity, based on six dimensions: general, management, business, market, technology and financial. The evaluations were discretized and used as input for a neural network model implemented with the Keras API from the Scikit-Learn library in Python. Results: The machine learning algorithm achieved an accuracy of 75%, enabling the identification of business failures that could lead to organizational vulnerability.
Limitations of the research: The proposed model is applicable only to companies where consultants trained in the IncubE methodology have performed a prior organizational diagnostic assessment. Practical implications: This study supports consultants and business managers to anticipate critical business situations and prevent the occurrence of failure scenarios. Social implications: Possible reduction of organizational problems and consequent reduction in the number of companies in crisis or bankrupt. Originality / value: There are few works in literature that deal with prediction of business vulnerabilities and none were found that use a methodology which evaluates the five organizational dimensions of the CERNE methodology: entrepreneur, technology, market, management and financial.