摘要
本論文的目標為大數據分析之財務金融應用，主要研究如何使用機器學習技術以開發高投資報酬率的智慧型股票選股策略。我們利用技術指標和財務指標等資料，這些指標是根據2014～2019年間，台灣的股票交易數據、公司的月報和財報所製作，並採用隨機森林模型預測股票的漲跌。此外並依預測結果的importances高低挑選features製成選股策略，最後再進行策略參數的優化。實驗結果顯示，利用得分較高的features製成的策略，獲利能高於用全部features所製成的策略，且參數優化後，獲利上升。採用我們的選股策略所得到的獲利高於同時間購買0050（ETF）的獲利。
並列摘要
This paper investigates a financial application of stocks for big data analysis. We aim to propose a smart stock picking strategy with high return-on-investment using machine learning. To validate the effect of the proposed method, we firstly use the stocks exchange data of companies in Taiwan from 2014 to 2019, also the technical indexes and financial indexes are produced and used for experiments. These two kinds of indexes are made according to companies' monthly reports and financial reports. Secondly, random forest model is applied to predict the stock prices and the features are selected according to the importances of the forecast results. Finally, we optimize the parameters for our proposed strategy. The experimental results showed that, by using the selected features, the profit obtained was much higher than that of the strategy based on all of the features. Moreover, after the parameter optimization, the profit of the strategy was promoted significantly and higher than that made by purchasing 0050(ETF) at the same period of time.