本研究以資產配置為主題建構理財機器人，首先為明確了解基金的投資風格與類別，利用分類模型針對每檔基金進行分類，基金的類別分別有：美國市場、歐洲市場、日本市場、亞洲非日本市場、新興市場、公司債、高收益債、新興市場債以及美國長期國債。接著進行大類資產配置，本研究以常見之大類指數進行投資組合的配置，並作為後續基金投資組合建構的基礎以及績效回測的比較對象(Benchmark)。最後則是進行基金挑選，本研究提供兩種基金挑選方法，分別是：優勢策略與XGBoost評分模型，前者利用過去歷史表現進行基金挑選；後者則是將基金的技術指標與市場的總體指標作為XGBoost訓練之特徵，每期針對每檔基金進行打分，並且挑選分數較高之基金，作為基金投資組合建構之成分。主要研究結果如下：
1. 優勢策略所建構之基金投資組合，在歷年績效上擁有較高的報酬率，但也須承受較高風險。
2. XGBoost評分模型建構之基金投資組合，在歷年報酬上雖沒有穩定勝過優勢策略，但其風險與優勢策略或是Benchmark相比，卻擁有較低的波動度，尤其在景氣不好的年份中，XGBoost評分模型所建構之基金投資組合擁有較高之報酬與較低之波動。
由於本研究所提出之XGBoost評分模型能根據市場趨勢建構波動較小的基金投資組合，因此適合應用於理財機器人之建構，協助投資人建構風險小且獲利高的基金投資組合。

This study constructs a Robo-Advisor with the theme of asset allocation. First, to clearly understand the category of the fund, the classification model is used to classify each fund. There are US market, european market, japanese market, asian non-Japan. Markets, emerging markets, corporate bonds, high-yield bonds, emerging market bonds, and long-term US Treasury bonds in fund categories. Then proceed with the allocation of index. This study uses common indices for asset allocation. The result of index allocation is used as the basis of the construction of the follow-up fund portfolio and the benchmark for backtesting. Finally, perform the fund selection by two fund selection methods: the Dominant Strategy and the XGBoost Scoring Model. The former selects funds by historical performance; the latter is the inputs of the technical indicators and the macro indicators for XGBoost training. Each fund will be scored every training, and the funds with higher scores are selected as the components of the fund portfolio.
The major empirical results are as follows:
1. The Dominant Strategy had higher return and risk in the historical performance.
2. Although the XGBoost Scoring Model has not stabilized over the Dominant Strategy in the historical performance, its risk has a lower volatility than the benchmark or the Dominant Strategy. Especially, the XGBoost Scoring Model had higher return and lower risk in the bear markets.
The XGBoost scoring model proposed by this study is suitable for the construction of Robo-Advisor because of its lower risk. It could help investors build a profitable and riskless fund portfolio.

論文審定書 i
摘 要 ii
Abstract iii
目 錄 iv
圖 次 vi
表 次 viii
第一章 緒論 1
第一節 研究背景動機 1
第二節 研究目的 2
第三節 研究架構 3
第二章 文獻探討 4
第一節 基金實證文獻 4
第二節 投資組合實證文獻 5
第三章 研究方法 7
第一節 實驗架構 7
第二節 資料說明與處理 10
第三節 基金分類 12
第四節 大類資產配置 14
第五節 優勢策略 16
第六節 XGBoost基金評分模型 17
第七節 績效指標 24
第四章 實證結果 26
第一節 研究樣本與期間說明 26
第二節 基金分類模型結果 27
第三節 大類資產配置績效 29
第四節 基金投資組合建構 35
第五章 結論與建議 46
第一節 結論 46
第二節 後續建議 47
參考文獻 48

1. 林文宏(1998)，「全球股票型基金績效及持續性之研究」，碩士論文，國立東華大學國際企業管理研究所。
2. 林明德(2012)，以投資績效評估指標探討基金的篩選方式與策略，碩士論文，國立政治大學經營管理碩士學程。
3. Breiman, L., Friedman, J. H., Olshen, R. A., and Stone, C. J. (1984). Classification and Regression Trees. London:Chapman & Hall.
4. Brinson, Gary P., Brian D. Singer & Gilbert L. Beebower. (1991). Determinants of Portfolio Performance II: An Update. Financial Analysts Journal, 47, 40-48.
5. Carhart, M. M. (1997). On Persistence In Mutual Fund Performance. Journal of Finance, 52(1), 57-82.
6. Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. In Proceedings of The 22Nd ACM SIGKDD International Conference On Knowledge Discovery and Data Mining, New York, USA: ACM, 785–794.
7. Friedman, Jerome H. (1999). Greedy function approximation: A gradient boosting machine. Ann. Statist. 29 (2001), no. 5, 1189-1232.
8. Goetzmann, W. N. & Ibbotson, R. G. (1994). Do winners repeat? Patterns in mutual fund performance. Journal of Portfolio Management, 20, 9-17.
9. Grubel, H. G.. (1968). Internationally Diversified Portfolios: Welfare Gains and Capital Flows. The American Economic Review,58(5), 1299–1314.
10. Levy, Haim & Sarnat, Marshall. (1970). International Diversification of Investment Portfolios. American Economic Review, American Economic Association, 60(4), pages 668-675, September.
11. Malkiel, Burton G. (1995). Returns from Investing in Equity Mutual Funds 1971-1991. Journal of Finance, 50(2), June 1995.
12. Markowitz, H. (1952). Portfolio selection. Journal of Finance, 7(1), 77-91.
13. Sharpe, W. F. (1966). Mutual fund performance. Journal of Business, 39, 119-138.
14. Sharpe, W. F. (1992). Asset allocation: Management style and performance measurement. Journal of Portfolio Management, 18(2), 7-19.
15. Tukey, J. W. (1977). Exploratory Data Analysis. Reading, Massachusetts: Addison-Wesley.
16. Wainscott, Craig B. (1990). The Stock-Bond Correlation and Its Implications for Asset Allocation. Financial Analysts Journal, 46(4), 55-79.