The Study of the ANN Models for Predicting the Sovereign Bond Default Swaps

Abstract

This study compares five Artificial Neural Network (ANN) models, such as Random Forests (RF), Support Vector Machine (SVM), Decision Tree, Naive Bayes Classifier (NBC), and K-Nearest Neighbor Algorithm (KNN), to predict the best model for sovereign bond default swaps (CDSs). We use the seven sentiment indicators, including the S&P500, VIX, USD index, LIBOR, Put/Call ratio, Commodity Research Bureau (CRB), and Association of Individual Investors (AAII). The result showed that the RF and Decision tree were the best prediction models for sovereign bond CDSs with higher accuracy and lower errors. These findings suggest that investors can use RF and Decision Tree models to set their future investment plans and minimize risk. Furthermore, they can use ANN models to forecast the CDS to build the hedging strategy when facing the downturn economic cycle.