臺灣博碩士論文加值系統

本研究提出了一種融合軟訊息的滯後多元貝葉斯結構GARCH模型，稱為SHMBS-
GARCH，用於描述不同經濟狀態下的多維金融時間序列動態。我們首先運用
De-GARCH 技術去除每條金融時間序列中的GARCH 效應。接著，構建一個針對
De-GARCH 時間序列的滯後多元貝葉斯結構模型，同時捕捉趨勢、季節性、循環
模式以及內生（或外生）協變量效應。特別的是，我們將從每日金融新聞中提取
的軟訊息納入模型的滯後部分，以反映經濟對時間序列行為的影響。為估計參數，
我們提出了一種MCMC 算法，而模擬研究表示，所提出的算法能夠獲得好的估
計結果。實證研究利用2016 年1 月至2020 年12 月期間的道瓊工業、納斯達克
和費城半導體指數數據，評估了所提出模型的性能。數值分析顯示，所提出的模
型在擬合和預測精度方面優於競爭模型。

This study proposes a hysteretic multivariate Bayesian structural GARCH model
integrating soft information, denoted by SH-MBS-GARCH, to describe
multidimensional financial time series dynamics under different economic states. We
first employ the De-GARCH technique to remove GARCH effects from each financial
time series. Next, we construct a hysteretic multivariate Bayesian structural model for
the De-GARCH time series, simultaneously capturing trends, seasonality, cyclic
patterns, and endogenous (or exogenous) covariate effects. In particular, we
incorporate soft information extracted from daily financial news into the model's
hysteretic part, reflecting economic influences on time-series behavior. An MCMC
algorithm is proposed for parameters estimation. Simulation studies reveal that the
proposed algorithm can obtain satisfactory estimation results. The empirical study
utilizes data from the Dow Jones Industrial, Nasdaq, and Philadelphia Semiconductor
indices spanning from January 2016 to December 2020 to evaluate the performance
of the proposed model. Numerical analysis demonstrates that the proposed model
outperforms competing models in terms of fitting and predictive accuracy.

摘要 ................................................................................................... i
Abstract ............................................................................................. ii
1. Introduction ................................................................................... 1
2. Literature Review ......................................................................... 2
2.1 Multivariate Bayesian structural models .....................................................2
2.2 Hysteretic time series models ......................................................................4
2.3 Soft information ..........................................................................................5
2.4 Principal component analysis ......................................................................6
3. Methodology ................................................................................ 6
3.1 The proposed SH-MBS-GARCH model ..........................................................7
3.2 Parameter estimation .................................................................................9
4. Simulation .................................................................................. 11
4.1 Simulation study .......................................................................................11
4.2 Parameter estimation and model performance .........................................13
5. Empirical Analysis ......................................................................... 17
5.1 Data description ........................................................................................17
5.2 Empirical study ..........................................................................................19
6. Discussion .................................................................................... 24
Reference ........................................................................................ 25
Appendix A ...................................................................................... 27
Appendix B ...................................................................................... 28
Appendix C ...................................................................................... 29

[1] Baker SR, Bloom N, Davis SJ. (2016). Measuring Economic Policy Uncertainty. The Quarterly Journal of Economics, 131, 1593-1636.
[2] Baker SR, Bloom N, Davis SJ, Terry SJ. (2020). Covid-induced economic uncertainty. NBER Working Paper No. 26983. National Bureau of Economic Research, Cambridge, MA.
[3] Bollerslev T. (1986). Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31, 307–327.
[4] Chen J, Ma F, Qiu X, Li T. (2023). The role of categorical EPU indices in predicting stock-market returns. International Review of Economics & Finance, 87, 365–378.
[5] Chen WS, Truong BC. (2016). On double hysteretic heteroskedastic model. Journal of Statistical Computation and Simulation, 86, 2684–2705.
[6] Cheng WH, Hung JC. (2011). Skewness and leptokurtosis in GARCH typed VaR estimation of petroleum and metal asset returns. Journal of Empirical Finance, 18, 160–173.
[7] Durbin J, Koopman SJ. (2002). A simple and efficient simulation smoother for state space time series analysis. Biometrika, 89, 603–616.
[8] Engle RF. (1982). Autoregressive conditional heteroskedasticity with estimates of the variance of United Kingdom inflation. Econometrica, 50, 987–1007.
[9] George EI, McCulloch RE. (1997). Approaches for Bayesian variable selection. Statistica Sinica, 339–373.
[10] Gentzkow M, Kelly B, Taddy M (2019). Text as data. Journal of Econometrics, 57(3):535–574.
[11] Gocke M. (2002). Various concepts of hysteresis applied in economics. Journal of Econometrics Surveys, 16, 167–88.
[12] Jammalamadaka SR, Qiu J, Ning N. (2019). Predicting a stock portfolio with the multivariate Bayesian structural time series model: do news or emotions matter? International Journal of Artificial Intelligence, 17, 81–104.
[13] Qiu J, Jammalamadaka SR, Ning N. (2018). Multivariate Bayesian structural time series model. Journal of Machine Learning Research, 19, 2744–2776.
[14] Li G, Guan B, Li WK, Yu PL. (2015). Hysteretic autoregressive time series models. Biometrika, 102, 717–723.
[15]
[16] Madigan D, Raftery AE. (1994). Model selection and accounting for model uncertainty in graphical models using Occam’s window. Journal of the American Statistical Association, 89, 1535–1546.
[17] Musto C, Semeraro G, Polignano M. (2014). A comparison of lexicon-based approaches for sentiment analysis of microblog posts. In Proceedings of the 8th International Workshop on Information Filtering and Retrieval - Workshop of the XIII AI*IA Symposium on Artificial Intelligence, 59-68.
[18] Petersen MA. (2004). Information: Hard and soft. Working Paper.
[19] Theodossiou P. (1998). Financial data and the skewed generalized t distribution. Managemet Science, 44, 1650–1661.
[20] Tsai MF, Wang CJ. (2017). On the risk prediction and analysis of soft information in finance reports. European Journal of Operational Research, 257, 243–250.
[21] Wu S, Chen R. (2007). Threshold variable determination and threshold variable driven switching autoregressive models. Statistica Sinica, 17, 241-–64.
[22] Yu J, Shi X, Guo D, Yang L. (2021). Economic policy uncertainty (EPU) and firm carbon emissions: evidence using a China provincial EPU index. Energy Economics, 94, 105071.22