Loading...
|
Please use this identifier to cite or link to this item:
https://nccur.lib.nccu.edu.tw/handle/140.119/158567
|
| Title: | 雙重機器學習於循序處方效果
Double Machine Learning for Sequential Treatment Effects |
| Authors: | 邢芳瑜
Hsing, Fang-Yu |
| Contributors: | 莊皓鈞
周彥君
Chuang, Hao-Jun
Chou, Yen-Chun
邢芳瑜
Hsing, Fang-Yu |
| Keywords: | 雙重機器學習
統計與因果推論
延滯處方效果
累積處方效果
Double Machine Learning
Statistical and Casual Inference
Lagged Treatment Effect
Cumulative Treatment Effect |
| Date: | 2024 |
| Issue Date: | 2025-08-04 14:25:25 (UTC+8) |
| Abstract: | 在不斷變化的商業環境下，企業做決策時需要評估各種策略或是處方帶來的效果，近年來因果推論研究引進了機器學習框架，其中雙重機器學習（Double Machine Learning, DML）是廣受歡迎的機器學習框架之一。過往DML在管理領域的文獻中，主要聚焦在橫截面資料的單期處方效果估計，然而許多實務情境中，決策需要考慮多期的處方效果，因為當期決策不僅影響當期結果，也可能對未來結果產生延遲影響。
因此，本研究第一部分著重在模擬兩期動態資料，以雙重機器學習估計當期和延滯處方效果。有鑒於決策的影響經常不是即時的，需要一段時間才能觀察到成效，因此第二部分著重於多期處方對累計結果變數的影響，模擬多期動態資料進行累積效果的估計，另外也透過銀行業的實證資料，分析一段時間的推播提醒是否影響顧客行為。
本研究結果具有重要的實務意涵，有助於企業在動態變化的環境中，基於多期處方效果更精準衡量策略效果，從而在資源配置與商業策略上做出調整。例如，企業可以透過區分當期效果與滯後效果來全面評估行銷活動的真實價值，或是根據每期係數的估計結果規劃最佳的行銷訊息推播時機，以最大化最終收益。
In the ever-evolving business environment, companies must evaluate the effectiveness of various strategies or treatments when making decisions. Recently, causal inference research has increasingly incorporated machine learning frameworks, among which Double Machine Learning (DML) has emerged as one of the most widely adopted approaches. Previous applications of DML in the management literature have primarily focused on estimating single-period treatment effects using cross-sectional data. However, in many real-world scenarios, decision-making requires consideration of multi-period treatment effects, as current decisions can influence not only immediate outcomes but also have delayed impacts on future results.
The first part of this study focuses on simulating two-period dynamic data and estimating both contemporary and lagged treatment effects using the DML framework. Given that the impact of decisions often appears over time rather than instantly, the second part of the study emphasizes the cumulative impact of multi-period treatments on outcome variables. This involves simulating multi-period dynamic data to estimate cumulative effects. In addition, the study utilizes empirical data from the banking industry to analyze whether a series of push notifications over time influences customer behavior.
The findings of this research hold significant practical implications, enabling businesses to more accurately assess strategic effectiveness based on multi-period treatment effects in a dynamic environment. For instance, by distinguishing between immediate and delayed effects, firms can comprehensively evaluate the true value of marketing activities. Moreover, using the estimated coefficients across periods, companies can better plan the optimal timing of marketing message delivery to maximize overall returns. |
| Reference: | 周平（2023）。統計機器學習：理論建構與因果分析﹝博士論文﹞。國立政治大學。臺灣博碩士論文知識加值系統。https://hdl.handle.net/11296/8725gk。
Bach, P., Chernozhukov, V., Kurz, M. S., & Spindler, M. (2022). DoubleML: An object-oriented implementation of double machine learning in Python. Journal of Machine Learning Research, 23(53), 1-6.
Bach, P., Kurz, M. S., Chernozhukov, V., Spindler, M., & Klaassen, S. (2024). DoubleML: An object-oriented implementation of double machine learning in R. Journal of Statistical Software, 108(3), 1–56. https://doi.org/10.18637/jss.v108.i03
Battocchi, K., Dillon, E., Hei, M., Lewis, G., Oprescu, M., & Syrgkanis, V. (2021). Estimating the long-term effects of novel treatments. Advances in Neural Information Processing Systems, 34, 2925-2935.
Battocchi, K., Dillon, E., Hei, M., Lewis, G., Oka, P., Oprescu, M., & Syrgkanis, V. (2019). EconML: A Python package for ML-based heterogeneous treatment effects estimation (Version 0.15.1). GitHub. Retrieved December 2, 2024, from https://github.com/py-why/EconML
Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W., & Robins, J. (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21(1), C1–C68.
Colangelo, K., & Lee, Y.Y. (2023). Double debiased machine learning nonparametric inference with continuous treatments. arXiv. https://arxiv.org/abs/2004.03036
Deng, Y., Zhang, X., Wang, T., Wang, L., Zhang, Y., Wang, X., ... & Peng, X. (2023). Alibaba realizes millions in cost savings through integrated demand forecasting, inventory management, price optimization, and product recommendations. INFORMS Journal on Applied Analytics, 53(1), 32–46. https://doi.org/10.1287/inte.2022.1145
Ellickson, P. B., Kar, W., & Reeder III, J. C. (2023). Estimating marketing component effects: Double machine learning from targeted digital promotions. Marketing Science, 42(4), 704-728.
Hünermund, P., Louw, B., & Caspi, I. (2023). Double machine learning and automated confounder selection: A cautionary tale. Journal of Causal Inference, 11(1), 20220078. https://doi.org/10.1515/jci-2022-0078
Lewis, G., & Syrgkanis, V. (2021). Double/Debiased Machine Learning for Dynamic Treatment Effects. Advances in Neural Information Processing Systems, 27, 22695–22707. Neural information processing systems foundation.
Lewis, G., & Syrgkanis, V. (2021). Double/Debiased Machine Learning for Dynamic Treatment Effects via g-Estimation Vasilis Syrgkanis. ArXiv.
Semenova, V., Goldman, M., Chernozhukov, V., & Taddy, M. (2023). Inference on heterogeneous treatment effects in high‐dimensional dynamic panels under weak dependence. Quantitative Economics, 14(2), 471-510.
Yang, J. C., Chuang, H. C., & Kuan, C. M. (2020). Double machine learning with gradient boosting and its application to the Big N audit quality effect. Journal of Econometrics, 216(1), 268-283. |
| Description: | 碩士
國立政治大學
資訊管理學系
112356001 |
| Source URI: | http://thesis.lib.nccu.edu.tw/record/#G0112356001 |
| Data Type: | thesis |
| Appears in Collections: | [資訊管理學系] 學位論文
|
Files in This Item:
| File |
Description |
Size | Format | |
|---|
| 600101.pdf | | 2192Kb | Adobe PDF | 0 | View/Open |
|
All items in 政大典藏 are protected by copyright, with all rights reserved.
|
