Affiliations: [a] Dpto de Informática, Universidad Carlos III de Madrid, Madrid, Spain
| [b] Dpto de Ingeniería de Sistemas Informáticos y Telemáticos, Universidad de Extremadura, Mérida, Spain
| [c] Dpto de Lenguajes y Ciencias de la Computación, Universidad de Málaga, Málaga, Spain
Correspondence:
[*]
Corresponding author. Rafael M. Luque Baena, Dpto de Lenguajes y Ciencias de la Computación, Universidad de Málaga, Málaga, Spain. E-mail: [email protected].
Abstract: Initial public offerings often show abnormal fist-day returns. These, usually referred to as underpricing, are difficult to predict. Among the main obstacles, we could mention challenges like the fact that not all relevant variables have been identified yet; the mix of weak and strong indicators or the prevalence of outliers. In this context, we suggest that adaptive neuro-fuzzy inference systems and fuzzy rule-based system with genetic optimization have a lot to bring to the table. We test the predictive performance of these on a sample of 866 US IPOs and we benchmark them against six fuzzy algorithms and a set of eight classic machine learning alternatives. We conclude that both fuzzy systems, especially the former should be seriously considered in this domain.
Keywords: Fuzzy rule-based system, ANFIS, initial public offering, underpricing
