Percorrer por autor "Loureiro, Francisco Soares da Cruz"
A mostrar 1 - 1 de 1
Resultados por página
Opções de ordenação
- Corporate bankruptcy prediction : can KMV-Merton model add value to support vector machines forecasts?Publication . Loureiro, Francisco Soares da Cruz; Silva, Nuno Ricardo Raimundo Rodrigues Marques daThis dissertation aims to assess if the output from the KMV-Merton model, the so-called distance to default, can contribute to the support vector machines model with the ultimate goal of better forecasting the bankruptcy of a company. The considered dataset covers 248 non-financial U.S. companies between 2000 and 2018. It was found evidence that the distance to default contributes, within a given range of variables considered, to a better F1-Score using both cross-validation and percentage ratio split. Additionally, the results show that the distance to default is a better predictor than a simpler market-based variable such as the debt-to-equity ratio. This suggests that the Merton-model setup per se is useful for default prediction. As expected, taking the F1-Score as a reference, the results also indicate that using company information a year prior to default provides better results than using data two years prior to default. Lastly, given the dataset used and the assumptions stated, this study is not conclusive regarding which out-of-sample evaluation method offers better results, the percentage ratio split, or the stratified K-fold cross-validation.