| DC 欄位 |
值 |
語言 |
| DC.contributor | 經濟學系 | zh_TW |
| DC.creator | 郭家瑞 | zh_TW |
| DC.creator | Jia-Ruei Guo | en_US |
| dc.date.accessioned | 2024-2-20T07:39:07Z | |
| dc.date.available | 2024-2-20T07:39:07Z | |
| dc.date.issued | 2024 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=110429016 | |
| dc.contributor.department | 經濟學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 股票報酬預測在經濟與財金領域是一個相當重要的議題，本文研究目的為運用25個美國與臺灣財金經濟預測因子，採用七種不同機器學習模型預測已除權除息之臺灣加權股價指數報酬，並運用樣本外R²值與Clark-West統計量檢定模型之顯著性，繪出模型預測臺灣股市報酬線圖，並觀察模型所預測之股市報酬與實際股市報酬間的差異性，找尋適合預測臺灣股價報酬之機器學習模型。
本文首先採用Fama-French三因子模型為基礎，運用t統計量方式，初步選出適合之預測因子個數，接著，使用彈性網絡法選擇重要預測因子。接下來，運用Dong et al. (2022) 所定義之平均預測報酬指標，計算並繪出不同期間之平均預測報酬指標，初步觀察顯示，當期間愈長，其平均預測報酬指標呈現愈平滑的趨勢；接著，使用七種不同機器學習模型計算預測臺灣加權指數股價報酬，實證結果得出在Clark-West 檢定顯著之模型中，平均預測法為本文之最佳機器學習模型，而除了Clark-West 檢定顯著模型之外，在R²值大於零之模型方面，簡單組合法為本文之較佳機器學習模型。 | zh_TW |
| dc.description.abstract | Predicting stock returns is a significant issue in the field of economics and finance. The purpose of this study is to utilize 25 financial and economic forecasting factors from the United States and Taiwan and employ seven different machine learning models to predict the returns of the Taiwan Weighted Stock Index after adjusting for dividends and stock splits. The out-of-sample R² value and Clark-West statistic are used to assess the significance of the models. The study then presents a graphical representation of the predicted stock market returns based on the models and observes the differences between the predicted returns and actual stock market returns, aiming to identify a suitable machine learning model for predicting Taiwan stock price returns.
In this paper, the Fama-French three-factor model is initially used, employing t-statistics to select the appropriate number of predictive factors. Subsequently, the elastic net method is applied to select the important predictive factors. Next, the average forecast return indicator defined by Dong et al. (2022) is calculated and plotted for different periods. Preliminary observations show that the average forecast return indicator becomes smoother as the period lengthens. Seven different machine learning models are then used to compute the predicted returns of the Taiwan Weighted Stock Index. The empirical results indicate that the average forecast method is the best machine learning model in terms of significant results from the Clark-West test. However, among the models with R² values greater than zero but not significant according to the Clark-West test, the simple combination method is a better machine learning model in this study. | en_US |
| DC.subject | 臺灣股價報酬預測 | zh_TW |
| DC.subject | Fama-French三因子模型 | zh_TW |
| DC.subject | Clark-West 統計檢定 | zh_TW |
| DC.subject | 機器學習模型 | zh_TW |
| DC.subject | Taiwan stock return prediction | en_US |
| DC.subject | Fama-French three-factor model | en_US |
| DC.subject | Clark-West statistic test | en_US |
| DC.subject | machine learning models | en_US |
| DC.title | 運用機器學習方法預測股價報酬 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |