在金融衍生性商品中，選擇權一直是一種重要的基礎性產品，因此選擇權定價一直學者研究的重點。近40年來選擇權發展中最重要的成果就是Black- Scholes 選擇權定價模型。然而由于該模型理想化的假設，導致它在真實定價過程中容易出現明顯的誤差。但是神經網路模型有著利用資料開始自我學習的特性，可以不用假設條件，單純由資料確定模型的結構和參數。
本文選取2008年到2018年的臺灣加權指數選擇權（TXO）日資料作爲研究對象，利用Python構建NN神經網路模型，將買權資料分爲買權完整資料、買權價內資料、買權價平資料、買權價外資料、買權上漲趨勢資料、買權下跌趨勢資料共6類資料。再加上賣權的6類資料，一共12大類資料。分別進行訓練模型。最後採用MSE、MAE兩種誤差指標來評價不同模型的預測精度。
最後發現NN神經網路模型的定價精度大多優于BS模型的期權定價效果。同時NN模型的價外選擇權資料的定價效果更精確，幷且按漲跌趨勢劃分後的選擇權資料定價效果也比完整資料的定價效果要好。

Option is a significant basic product in financial derivatives. How to price an option is a major issue to many scholars. During the last 40 years, Black-Scholes option pricing model has been considered as the crucial research achievement. However, obvious bias occurs in the real market pricing procedure due to the idealized assumption of this model. The neural network model has the characteristic of using data to start self-learning, so the structure and parameters of the model can be determined by data without assuming conditions.
This thesis took TXO(2008-2018) as a research object, and used the Python to structure Neural Network(NN) model. Then the data of call option have been divided into 6 types , including‘all data’ ,‘in-the-price data’, ‘at-the-price data’, ‘out-the-price data’, ‘up-trend data’ and‘down-trend data’. The same classification is applied to the put option data. A total of 12 types of data have been trained by NN model separately. Finally, MSE and MAE are used to evaluate the accuracy of the forecasts of different models.
In conclusion, the pricing accuracy of the neural network model is substantially better than that of the Black-Scholes model. Meanwhile , the pricing effect of out-the-price option data is more accurate, and the pricing of up-trend option data has a good effect either.

[1] 李沃墻.(2000).台股重設型權證的評價績效比較陰.真理財金學報,91-112
[2] 周大鵬. (2008). 基於B-P神經網路的期權定價研究.(Doctoral dissertation,中國人民大學).
[3] 馬發強. (2012). 基於RBF神經網路的期權定價研究.(Doctoral dissertation, 中南大學).
[4] 張鴻彥, & 林輝. (2007). 基于小波神經網絡的期權定價模型. 東南大學學報 (自然科學版), 37(4), 716-720.
[5] 董瑩, 烏日嘎, & 齊淑華. (2013). 基於bp神經網路的期權定價模型. 魯東大學學報(自然科學版), 29(3), 196-199.
[6] 劉志强. (2005). 基于神經網路的期權定價模型. (Doctoral dissertation, 重慶大學).
[7] 劉旭彬. (2011). 基於神經網路方法的期權定價應用研究. (Doctoral dissertation, 暨南大學).
[8] 譚朵朵. (2008). 基於bp神經網路的s&p500指數期權定價. 統計與資訊理論壇, 23(11), 40-43.
[9] Amilon, H. (2003). A neural network versus black– scholes: a comparisonof pricing and hedging performances.Journal of Forecasting, 22(4), 317-335.
[10] Gençay, R., & Qi, M. (2001). Pricing and hedging derivative securities with neural networks: bayesian regularization, early stopping, and bagging. IEEE Trans Neural Netw, 12(4), 726-734.
[11] Hinton, G. E. (2012). A practical guide to training restricted Boltzmann machines. In Neural networks: Tricks of the trade(pp. 599-619). Springer, Berlin, Heidelberg.
[12] Huang, S. C., & Wu, T. K. (2006, September). A hybrid unscented Kalman filter and support vector machine model in option price forecasting. In International Conference on Natural Computation (pp. 303-312). Springer, Berlin, Heidelberg.
[13] Hinton, G. E., Osindero, S., & Teh, Y. W. (2006). A fast learning algorithm for deep belief nets. Neural computation, 18(7), 1527-1554.
[14] Hutchinson, J. M., Lo, A. W., & Poggio, T. (1994). A nonparametric approach to pricing and hedging derivative securities via learning networks. Journal of Finance, 49(3), 851-889.
[15] Liang, X., Zhang, H., Xiao, J., & Chen, Y. (2009). Improving option price forecasts with neural networks and support vector regressions. Neurocomputing, 72(13), 3055-3065.
[16] Panayiotis, A. C., Spiros, M. H., & Chris, C. (2004, July). Option pricing and trading with artificial neural networks and advanced parametric models with implied parameters. In Neural Networks, 2004.
proceedings. 2004 IEEE International Joint Conference on (Vol. 4, pp. 2741-2746). IEEE.
[17] Park, H., Kim, N., & Lee, J. (2014). Parametric models and non-parametric machine learning models for predicting option prices: empirical comparison study over kospi 200 index options. Expert Systems with Applications, 41(11), 5227-5237.
[18] Paul R. Lajbcygier, & Jerome T. Connor. (1997). Improved option pricing using artificial neural networks and, bootstrap methods. International Journal of Neural Systems, 8(04), 457-471.
[19] Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. nature, 323(6088), 533.
[20] Srivastava, R. K., Greff, K., & Schmidhuber, J. (2015). Highway networks. arXiv preprintarXiv:1505.00387.
[21] Wang, Y. H. (2009). Nonlinear neural network forecasting model for stock index option price: Hybrid GJR–GARCH approach. Expert Systems with Applications, 36(1), 564-570.
[22] Wu, S., Zhong, S., & Liu, Y. (2018). Deep residual learning for image steganalysis. Multimedia tools and applications, 77(9), 10437-10453.