| Summary: | 碩士 === 東吳大學 === 資訊管理學系 === 107 === With the innovation of financial products and the globalization of the financial market, it is hard to make appropriate choices for investors due to the information of the financial market is changing quickly. Machine learning is an excellent method to tackle various and tremendous datasets efficiently. The study attributes 11 key factors which may affect the performance and risk of mutual funds. By means of combining the algorithm of K-means and support vector machine (SVM) which is called K-SVM model, the fund datasets are able to be clarified properly according to those similar attributions.
In the first phase, the training dataset was divided into four groups by using K-means algorithm. In order to select outperformed funds, we applied SVM algorithm to classify the dataset of each group separated by K-means already. According to the result of the study, accuracy rates of SVM and K-SVM model were both over 95%. In comparison to the single SVM model, K-SVM models presented much more outstanding performance. By way of importing several test datasets, most accuracy rates of K-SVM models were outperformed in small datasets while the performance of signal SVM model was good and stable instead.
|
|---|
