A Study on Securities Selection using a Stepwise Machine Learning Approach

This study seeks to construct a portfolio using a stepwise machine learning methodology and to demonstrate its potential applicability in the stock market. This study used market trading information from 1,632 listed companies in the Korean stock market, extracted from FnGuide’s database over a five-year period.In the first step, the support vector machine (SVM) technique was employed to predict volatility errors in the daily trading data of individual stocks. In the second step, K-means clustering was used to construct 12 portfolios, whose effectiveness was evaluated by examining their short- and long-term investment performance. Since the research method proposed in this study can be extended to stock trading information from public financial datasets, it is expected to promote further research on the effectiveness of portfolios generated through a stepwise machine learning approach using readily accessible general stock trading data.