多因子模型相關的研究大多針對使用各種方法找出有效的因子組合，然而鮮少研究
接續探討多因子模型結合技術面策略之相關驗證，以及如何建置這些自動化回測與驗證
平台。對此本研究參考謝昀峻 (2018)提出之類似實驗設計為概念，以此為基礎加入多因
子模型結合技術面策略動態換股策略，同時改進了能夠根據條件動態配置資金權重，並
且以 Pardo (2011)所提出投組最佳化方法——移動窗格設計系統流程，使能夠動態驗證
策略穩定性，然而其系統架構與流程設計存在種種限制無法兼容上述需求，故本研究自
行以 Python 開發一套具備上述功能之自動化回測與驗證平台，此外由於移動窗格之系
統流程設計需要較多的運算資源，故參考林泓志 (2020)研究中使用的多節點偵測任務分
散式演算法設計系統架構，以多台主機同時運算分擔運算負載。
本研究以 8 種單因子與 4 種雙因子組合搭配各種策略組合進行實驗，主要交易策略
共有買入持有與動態換股 2 種策略，另外動態換股包含其他設定，例如最佳化參數窗格
配置中的固定參數、定錨、非定錨；資金權重分配方法中的等資金權重、最大化資金利
用方法。並且會再實驗以各種組合選擇不同群組之候選股以及持有不同最大持有股數之
差異，最後將績效以多種視覺化圖表呈現，從不同角度剖析策略表現。

Most of the researches related to multi-factor models are aimed at finding effective factor
combinations using various methods, but few studies continue to explore the verification of
multi-factor models combined with technical analysis strategies, and how to build these
automated backtesting and verification platforms. In this regard, this research refers to the
concept of similar experimental design proposed by 謝昀峻 (2018). Based on this, it adds a
multi-factor model combined with a technical analysis strategy dynamic exchange stock
strategy. At the same time, it improves the ability to dynamically allocate capital weights
according to conditions, and use walk forward analysis which is a portfolio optimization method,
proposed by Pardo (2011) to design system process, so let it can dynamically verify the stability
of the strategy. However, its system architecture and process design have various limitations
and cannot be compatible with the above requirements. Therefore, this research uses Python to
develop an automated backtest and verification platform with the above functions. In addition,
because the system process design of the walk forward analysis requires more computing
resources, the system architecture is designed with reference to the multi-node detection task
distributed algorithm used in the research of 林泓志 (2020). It uses multiple hosts to perform
simultaneous operations to share the computational load.
This study uses 8 single-factor and 4 dual-factor combinations with various strategy
combinations to conduct experiments. The main trading strategies include two strategies: buyand-hold and dynamic stock exchange. In addition, dynamic stock exchange strategies
includes other settings, such as in the optimization parameter window configuration, there are
fixed parameters, anchored, and non-anchored; There are equal capital weight and maximum
capital utilization methods in the capital weight distribution method. Besides, we will
experiment with various combinations to select candidate stocks in different groups and the
difference in holding different maximum number of shares. Finally, the performance will be
presented in a variety of visual charts to analyze the performance of the strategy from different
perspectives.

Carhart, M. M. (1997). On persistence in mutual fund performance. The Journal of Finance, 52(1), 57-82.
Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of financial economics, 33, 3-56.
Lev, B., & Srivastava, A. (2019). Explaining the recent failure of value investing. NYU Stern School of Business.
Li, H., Zhang, X., Li, Z., & Zheng, C. (2020). Overview of Machine Learning for Stock Selection Based on Multi-Factor Models. Paper presented at the E3S Web of Conferences.
McLean, R. D., & Pontiff, J. (2016). Does academic research destroy stock return predictability? The Journal of Finance, 71(1), 5-32.
Novy-Marx, R. (2013). The other side of value: The gross profitability premium. Journal of financial economics, 108(1), 1-28.
Pardo, R. (2011). The evaluation and optimization of trading strategies (Vol. 314): John Wiley & Sons.
Rao, A., & Gupta, A. (2019). Value investing is down. But is it out? , from https://www.msci.com/www/blog-posts/value-investing-is-down-but-is/01619618276
Ross, S. (1976). The arbitrage theory of capital asset pricing, ʽJournal of Economic Theoryʼ: Elsevier, Amsterdam.
Tortoriello, R. (2009). Quantitative strategies for achieving alpha: McGraw Hill.
Trombetta, G. (no date). Persistenza, Validazione e Walk Forward Analysis. from https://www.gandalfproject.com/index.php/articles/219-persistenza-validazione-e-walk-forward-analysis?fbclid=IwAR104AoRKagoRf8KGtWhUMxkFqpEYMEsbGg56X8U05wrhLLswvGGaERzVEk
Wendong, Y., Zhengzheng, L., & Bo, J. (2017). A multi-factor analysis model of quantitative investment based on GA and SVM. Paper presented at the 2017 2nd International Conference on Image, Vision and Computing (ICIVC).
丁鹏. (2016). 量化投资: 策略与技术: 电子工业出版社.
石川. (2019). Anomalies, Factors, and Multi-Factor Models. from https://mp.weixin.qq.com/s?__biz=MzIyMDEwNDk1Mg==&mid=2650878746&idx=1&sn=1473b1b862ffa4861b98fe8c90521807&scene=21#wechat_redirect
朱家泓. (2017). 做對5個實戰步驟 你就是賺錢高手. 台灣: 金尉股份有限公司.
林泓志（2020）。多商品相關性獲利穩定度分析與比較。未出版之碩士論文，國立中央大學資訊管理學系，桃園縣。
蔡永丞（2017）。初始興櫃認購價格的特徵因素與IFRS實施前後的影響分析。未出版之碩士論文，國立中央大學產業經濟研究所在職專班。
鄭皓元（2018）。以分散式運算分析多商品主副策略最適性之自動化平台設計與驗證。未出版之碩士論文，國立中央大學資訊管理學系，桃園縣。
謝昀峻（2018）。運用等分法與核心交易策略於台灣股票之自動化平台設計與實證研究。未出版之碩士論文，國立中央大學資訊管理學系，桃園縣。