学習期間が異なる株価予測機械学習モデルのアンサンブル学習による投資戦略の構築

抄録

The application of machine learning method to securitiesreturn prediction has been actively researched. However, areview of previous studies reveals that the learning periodvaries greatly depending on the report. Also, in practice,determining the learning period is often a challenge. Inthis study, we propose a new method to solve this problemby using an ensemble of models that differ only in theirtraining windows. To demonstrate the effectiveness of theproposed method, we compared test loss and stock priceprediction capabilities of machine learning models withvarious learning periods and their ensemble models.