過去四十幾年來,由於全球經濟詭譎多變,造成各國金融市場變化多端。各投資機構為降低投資風險,創造出許多避險工具,其中信用違約交換(Credit Default Swap)即為近年盛行的衡量風險工具之一,其係透過各式各樣的投資組合,進而有效率的將非系統性風險加以移轉。因此各金融機構在作投資決策時,常使用CDS作為重要的避險工具。 此外,近年來,人工智能及機械式學習也都被認為是未來金融市場的趨勢之一。因此本研究特順應此趨勢,採用人工智能神經模型,結合CDS避險工具,進一步預測希臘、愛爾蘭、義大利、葡萄牙以及西班牙等歐洲五國之CDS價格,研究期間自2008年1月1日至2015年12月31日,研究方法係採用倒傳遞類神經模型以及回饋式類神經模型等兩種模型進行探討。 實證結果顯示,回饋式類神經模型的預測績效優於倒傳遞類神經模型,而在國家分類中,本研究發現希臘在兩個模型的測試下績效值更優於愛爾蘭、義大利、葡萄牙以及西班牙等四個國家。
Since the global economy and various international financial markets have been experienced financial tsunami over the past two decades. Most investment institutions reduce their investment risk by trading various hedging risk instruments. The credit default swap (CDS) is one of the popular risk measurement tools, which has been applied to assort investment portfolios, and transfers unsystematic risk to speculators in an effective way. Therefore, various financial institutions usually employ CDS as an important hedging instrument. In recent years, artificial intelligence and machine learning have also been recognized as one of the trends of the future financial markets. This study complies with this trend and uses the back-propagation and recurrent neural networks to forecast the price of CDS for five European countries, including Greece, Ireland, Italy, Portugal and Spain. The sample period runs from January 1, 2008 to December 31, 2015. Empirical results show that the recurrent neural network model has better forecasting effect than back-propagation neural network. Additionally, Greece outperforms Ireland, Italy, Portugal and Spain based on these two neural networks model.