研究生(外文):Liu,Chia-Ling
論文名稱:利用大數據與智慧技術分析台灣債券型及固定收益ETF價格之研究
論文名稱(外文):A Study on Analyzing the Prices of Taiwan Bond and Fixed-Income ETFs Using Big Data and Intelligent Technologies
指導教授:劉柏伸
指導教授(外文):Po-Shen Liu
學位類別:碩士
校院名稱:僑光科技大學
系所名稱:資訊科技研究所
論文出版年:2025
語文別:中文
論文頁數:87
論文摘要
本研究針對台灣證券公司發行的 ETF 產品進行深入風險分析，由於美國聯儲的加息政策對全球經濟的深遠影響，加上國際市場的不確定性增加與投資人的風險意識提高，本研究將結合機器學習的智慧技術，通過數據分析來捕捉市場趨勢，並提高風險管理的適應性。
研究中結合決策樹和隨機森林模型，並透過SHAP分析辨識影響ETF價格的重要特徵。包括：透過SHAP數值進行模型可解釋性分析；引入K-means聚類分析以識別高收益潛力ETF群體；利用GridSearchCV優化隨機森林模型參數，以提升模型的預測效能。研究結果顯示，開盤價和成交量對價格預測具顯著影響，隨機森林模型優化後在測試資料集中達到準確率76%、F1分數70%。此外，透過K-means聚類分析，將ETF分為三大類群，以提供不同風險偏好投資者的策略參考。
本研究結果有助於提升投資決策的精確性及風險管理的有效性。最後，研究希望提供具體建議，以增強投資者在不斷變化的金融環境中做出決策的能力。
論文外文摘要
This study conducts an in-depth risk analysis of ETF products issued by securities firms in Taiwan. In response to the far-reaching impact of the U.S. Federal Reserve's interest rate hikes on the global economy—alongside increasing uncertainty in international markets and heightened investor risk awareness—this research integrates intelligent machine learning techniques with data analytics to capture market trends and enhance adaptability in risk management.
The study integrates Decision Tree and Random Forest models and employs SHAP analysis to identify key features influencing ETF prices. The methodology includes: interpreting model outputs through SHAP values to enhance explainability; applying K-means clustering to identify ETF groups with high return potential; and using GridSearchCV to optimize the hyperparameters of the Random Forest model to improve predictive performance. The findings indicate that opening price and trading volume significantly impact price prediction. The optimized Random Forest model achieved 76% accuracy and an F1 score of 70% on the test dataset. In addition, K-means clustering segmented the ETFs into three major groups, offering strategic references for investors with different risk preferences.
The results of this study contribute to improving the precision of investment decision-making and the effectiveness of risk management. Ultimately, the research aims to provide practical recommendations to enhance investors’ ability to make informed decisions in an ever-evolving financial environment.
