本研究旨在建立「證券經紀商同策略網絡」(Broker Same Strategies Network)、「股票報酬相關性網絡」(Stock Return Correlation Network)、及首創的「前後期股票報酬相關性網絡」(Lead-Lag Stock Return Correlation Network)，探討網絡參數與個股特性之關係後，以機器學習預測個股的極端漲跌情況，並觀察網絡結構是否能提升預測力，及參數是否存在對極端漲、跌情況的非對稱性。透過社會網絡分析(Social Network Analysis)之方法，本研究擬以股票網絡反映市場結構變化，以前後期股票網絡區分股票於市場中領先落後地位，並以經紀商網絡直接地刻劃投資人心理及行為，萃取其中網絡參數，例如中心化程度(Centralization)、模組化程度(Modularity)等變化，探討其與報酬、波動度、及極端情況之關聯。本研究發現，股票網絡呈現高Degree Centralization且低Modularity時市場趨於動盪，且Leading Degree Centrality顯示位居下游、市值較大的公司常是市場上的領先者，而經紀商網絡中Eigen Centralization與Modularity分別刻劃投資人資訊不對稱程度與羊群效應。
關於網絡結構用於極端報酬情況的預測，本文分別採用多元邏輯斯迴歸與機器學習模型(Xgboost)來比較兩者預測能力。結果顯示，多元邏輯斯迴歸無法有效反應網絡參數，而Xgboost的預測力則有顯著提升。此外，中心化程度(Degree centralization)、落後中心性(Lagging degree centrality)、及領先中心性(Leading degree centrality)位居重要變數前三名，而經紀商網絡參數則位居中後段，因此我認為金融市場上整體的共變性對於個股極端報酬情況的影響力遠大於個股內投資人行為的影響力，且股票在市場中屬於跟隨者或資訊不對稱程度越低時有較好的預測表現。最後實證發現，加入網絡參數使得極端下跌情況之AUC顯著高於極端上漲情況之AUC。上述結果顯示，網絡結構確實能夠預測極端報酬情況，並在極端下跌時有較好的效果。

In this study, I construct “Broker Same Strategies Network”, “Stock Return Correlation Network”, and an innovative “Lead-Lag Stock Return Correlation Network” to explore the relations between network parameters and the characteristics of stocks, and to predict the extreme stock return condition with machine learning, and test whether network structure parameters can enhance the predictability. Also, I test whether there is an asymmetric effect between network structures and extreme returns. Through social network topology, I attempt to reflect the market changes by stock network, distinct leaders or followers in the market by lead-lag stock network, and capture investor behavior by broker network directly. I extract parameters, such as centralization and modularity, from networks to study their relations with return, volatility, and extreme stock return conditions. This study shows that high degree centralization with low modularity in the stock network reflects a volatile market condition; leading degree centrality reflects that downstream firms and large size firms are usually leaders in the market; eigen centralization and modularity in the broker network can reflect information asymmetry level and herding effect respectively.
This study employs multinomial logistic regression and machine learning model (Xgboost) to predict the extreme condition and compare the predictability of them. Result shows that multinomial logistic regression cannot make good use of network parameters, but Xgboost, however, has a significant improvement. Besides, degree centralization, lagging degree centrality, and leading degree centrality are the top 3 important variables. On the other hand, broker network parameters are in the middle rank. Thus, I think the co-movement of stocks in the financial market has greater influence on extreme return conditions than investors’ behavior. In addition, predictability is better when a stock is follower in the market and when the level of information asymmetry is lower. Finally, AUC of downward extreme return condition is significant higher than that of upward extreme return condition after I put network parameters into the model. In general, network structure does enhance the prediction and is more powerful on the downward extreme condition.

[論文審定書+i]
[摘要+ii]
[ABSTRACT+iv]
[CONTENTS+vi]
[LIST OF FIGURES+viii]
[LIST OF TABLES+ix]
[Chapter 1 INTRODUCTION+1]
[Chapter 2 LITERATURE REVIEW+5]
[2.1 Behavioral Finance+5]
[2.2 Network Analysis+6]
[2.2.1 Network constructions in finance+6]
[2.2.2 Application of network analysis in finance+7]
[Chapter 3 METHODOLOGY+10]
[3.1 Data Description+10]
[3.2 Extreme Return Condition+11]
[3.3 Network Construction+12]
[3.3.1 Stock Return Correlation Network+12]
[3.3.2 Lead-Lag Stock Return Correlation Network+17]
[3.3.3 Broker Same Strategy Network+20]
[3.3.4 Machine Learning+24]
[Chapter 4 EMPIRICAL RESULTS+26]
[4.1 Descriptive Statistics+26]
[4.1.1 Stock Market+26]
[4.1.2 Extreme Return Condition+27]
[4.1.3 Stock Return Correlation Network+28]
[4.1.4 Lead-Lag Stock Return Correlation Network+30]
[4.1.5 Broker Same Strategy Network+35]
[4.2 The Relationship between Market and Networks+37]
[4.3 Extreme Stock Return Prediction+45]
[4.3.1 Multinomial Logistic Regression+45]
[4.3.2 Extreme Gradient Boosting+51]
[Chapter 5 CONCLUSION+57]
[REFERENCES+59]
[APPENDIX+66]

Abreu, D., & Brunnermeier, M. K. (2003). Bubbles and crashes. Econometrica, 71(1), 173-204.
Albert, R., & Barabási, A. L. (2002). Statistical mechanics of complex networks. Reviews of modern physics, 74(1), 47.
Avramov, D., Chordia, T., & Goyal, A. (2006). The impact of trades on daily volatility. The Review of Financial Studies, 19(4), 1241-1277.
Baker, M., & Wurgler, J. (2006). Investor sentiment and the cross‐section of stock returns. The Journal of Finance, 61(4), 1645-1680.
Bakshy, E., Rosenn, I., Marlow, C., & Adamic, L. (2012, April). The role of social networks in information diffusion. In Proceedings of the 21st international conference on World Wide Web (pp. 519-528). ACM.
Bikhchandani, S., Hirshleifer, D., & Welch, I. (1992). A theory of fads, fashion, custom, and cultural change as informational cascades. Journal of political Economy, 100(5), 992-1026.
Blasco, N., Corredor, P., & Ferreruela, S. (2012). Does herding affect volatility? Implications for the Spanish stock market. Quantitative Finance, 12(2), 311-327.
Boss, M., Elsinger, H., Summer, M., & Thurner 4, S. (2004). Network topology of the interbank market. Quantitative finance, 4(6), 677-684.
Campbell, J. Y., & Kyle, A. S. (1993). Smart money, noise trading and stock price behaviour. The Review of Economic Studies, 60(1), 1-34.
Caparrelli, F., D'Arcangelis, A. M., & Cassuto, A. (2004). Herding in the Italian stock market: a case of behavioral finance. The Journal of Behavioral Finance, 5(4), 222-230.
Christie, W. G., & Huang, R. D. (1995). Following the pied piper: Do individual returns herd around the market?. Financial Analysts Journal, 51(4), 31-37.
Colla, P., & Mele, A. (2009). Information linkages and correlated trading. The Review of Financial Studies, 23(1), 203-246.
Donner, R. V., Zou, Y., Donges, J. F., Marwan, N., & Kurths, J. (2010). Recurrence networks—a novel paradigm for nonlinear time series analysis. New Journal of Physics, 12(3), 033025.
Fawcett, T. (2006). An introduction to ROC analysis. Pattern recognition letters, 27(8), 861-874.
Garfinkel, J. A., & Sokobin, J. (2006). Volume, opinion divergence, and returns: A study of post–earnings announcement drift. Journal of Accounting Research, 44(1), 85-112.
Heiberger, R. H. (2014). Stock network stability in times of crisis. Physica A: Statistical Mechanics and its Applications, 393, 376-381.
Hein, O., Schwind, M., & Spiwoks, M. (2012). Network centrality and stock market volatility: The impact of communication topologies on prices. Journal of Finance and Investment Analysis, 1(1), 199-232.
Hochberg, Y. V., Ljungqvist, A., & Lu, Y. (2007). Whom you know matters: Venture capital networks and investment performance. The Journal of Finance, 62(1), 251-301.
Kumar, M., & Thenmozhi, M. (2006). Forecasting stock index movement: A comparison of support vector machines and random forest.
Le Bris, D., & Hautcoeur, P. C. (2010). A challenge to triumphant optimists? A blue chips index for the Paris stock exchange, 1854–2007. Financial History Review, 17(2), 141-183.
Marwan, N., Donges, J. F., Zou, Y., Donner, R. V., & Kurths, J. (2009). Complex network approach for recurrence analysis of time series. Physics Letters A, 373(46), 4246-4254.
Menardi, G., & Torelli, N. (2014). Training and assessing classification rules with imbalanced data. Data Mining and Knowledge Discovery, 28(1), 92-122.
Nagelkerke, N. J. (1991). A note on a general definition of the coefficient of determination. Biometrika, 78(3), 691-692.
Newman, M. E. (2003). The structure and function of complex networks. SIAM review, 45(2), 167-256.
Nobi, A., Maeng, S. E., Ha, G. G., & Lee, J. W. (2014). Effects of global financial crisis on network structure in a local stock market. Physica A: Statistical Mechanics and its Applications, 407, 135-143.
Onnela, J. P., Chakraborti, A., Kaski, K., Kertesz, J., & Kanto, A. (2003). Dynamics of market correlations: Taxonomy and portfolio analysis. Physical Review E, 68(5), 056110.
Ozsoylev, H. N., Walden, J., Yavuz, M. D., & Bildik, R. (2013). Investor networks in the stock market. The Review of Financial Studies, 27(5), 1323-1366.
Patel, J., Shah, S., Thakkar, P., & Kotecha, K. (2015). Predicting stock and stock price index movement using trend deterministic data preparation and machine learning techniques. Expert Systems with Applications, 42(1), 259-268.
Patel, S. A., & Sarkar, A. (1998). Crises in developed and emerging stock markets. Financial Analysts Journal, 54(6), 50-61.
Serrano, M. A., & Boguná, M. (2003). Topology of the world trade web. Physical Review E, 68(1), 015101.
Shiller, R. J. (1987). Investor behavior in the October 1987 stock market crash: Survey evidence. NBER Working Paper No. 2446.
Sias, R. W. (2004). Institutional herding. The Review of Financial Studies, 17(1), 165-206.
Siegel, J. J. (1992). Equity risk premia, corporate profit forecasts, and investor sentiment around the stock crash of October 1987. Journal of Business, 557-570.
Tumminello, M., Di Matteo, T., Aste, T., & Mantegna, R. N. (2007). Correlation based networks of equity returns sampled at different time horizons. The European Physical Journal B, 55(2), 209-217.
Yao, Z., Zhang, Y., Lin, L., Zhou, Y., Xu, C., Jiang, T., & Alzheimer's Disease Neuroimaging Initiative. (2010). Abnormal cortical networks in mild cognitive impairment and Alzheimer's disease. PLoS computational biology, 6(11), e1001006.