A Comparison of the Particle Swarm Optimization-Support Vector and Cross Entropy-Support Vector Machines in Predicting Financial Distress

摘要

Financial distress is a condition that refers to a declining stage in the financial condition of a company that would happen before the company goes bankrupt. The competence in predicting financial distress becomes an important research due to the advantage in preventing companies' financial failure. Moreover, a financial distress prediction model will benefit the investors and creditors. This research develops a financial distress prediction model for listed manufacturing companies in Indonesia using a Support Vector Machine (SVM). Mathematically, an SVM is formulated in the form of quadratic programming, which findsan optimal solution of quadratic programming that is not easy and time consuming. In this research, Particle Swarm Optimization (PSO) and Cross Entropy (CE) are used to optimize one of the SVM's parameters -known as Lagrange multipliers - to find the optimal solution or near optimal solution of the dual Lagrange SVM. The accuracy of the prediction model and computation time will be compared between standard SVM, PSO-SVM and CE-SVM. The results of the experiment show that the accuracy of the model using PSO-SVM and CE-SVM is comparable with a shorter computational time compared to standard SVM.