랜덤 포레스트 모델을 활용한 채권형펀드 성과 예측 연구

When choosing financial investment products, investors struggle to construct an optimal portfolio. In this study, we propose a machine learning model that predicts the relative performance of fixed income funds one month out in order to help investors make portfolio choices. To verify the predictive power, we build a random forest model using performance data on fixed income funds from December 1, 2017, to January 31, 2023, as well as macro indicator data such as government bond interest rates, the won-to-US-dollar exchange rate, the domestic growth rate, and the consumer price index. The results of this study are as follows. First, the prediction results show accuracy and precision of more than 90%. Second, prediction power increases when characteristics are selected using BorutaPy, and third, cross-validation confirms the stability of the model. While most previous academic studies directly predict the price and return on investment from individual financial instruments, this study utilizes a classifier to guide the selection of relatively good performance groups. In this respect, it differs from previous studies. Therefore, this study is expected to indirectly help investors make investment decisions about fixed income funds. For future research, it is necessary to develop an optimal prediction model and to improve accuracy by comparing various machine learning models, such as deep learning and the support vector machine.