University of California, San Diego

 Names : Po-Ya Hsu Kuang-Hsuan LeeTitle: Support Vector Regression for Financial Market ForecastingOutline1. Motivation:People desire to predict the market prices with historical information or fundamental information from news. However, both parts are difficult to predict market prices. With the maturity of artificial intelligence (AI) in recent years, we are equipped with some tools to make better and reasonable predictions. We are interested in making big money and are eager to explore how the prices in financial market flow with time.2. Statement of problemAs for the market price prediction with market data:The three challenging issues in financial time series processing are noise, non-linearity and non-stationarity.Traditional methods such as Autoregressive (AR) model and the autoregressive integrated moving average (ARIMA) model do not work because they are good at handling linear and stationary time series.Although artificial neural networks (ANN), which minimize empirical risk principle in its learning process, can handle this situation, it also has problems about local minimum traps and the difficulty to decide hidden layer size and learning rate.SVR is better than ANN because it uses structural risk minimization principle which considers both the training error and the capacity of the regression model to minimize the upper bound of generalization error. The main problem with SVR is that we require practitioner experience to determine its hyperparameters and kernel functions.Recently, chaotic firefly algorithm for optimizing the SVR hyperparameters is good for finding hyperparameters for SVR.However, predicting a market stock price is not enough with only its historical information. It can also be impacted by other markets’ prices. For example, A, B are competitors, and there exists some kind of relationship between A and B. For instance, A is IT industry, and B is food industry. People usually transfer money from one to another due to the global market risk. As a result, we want to add clustering features to improve the model.Besides, most papers do not include news information and we think it is also important. Along with the powerful growth in NLP, we can find the valuable feature from news more easily. In finance, it tends to be fundamental analysis. However, the characteristic of company has different impact on trade-off between information from news and information from markets. We will consider its characteristic to weigh the weights of them.Moreover, we would like to extract the seasonal features from market. 3. Wishlists:Data Intel and Microsoft (9/12/2007- 11/11/2010 & ) and National Bank shares (6/27/2008- 8/29/2011)Yahoo Finance Data from NYSEFinancial newsMethod : Step 1 : train model A with SVR and chaos-based firefly algorithmStep 2 : train model B with SVM with features from financial news. Step 3 : combine A,B, the clustering features, and the the characteristic of company into SVM model to do prediction (SVR+ Model Predicitve Control (MPC)) Innovative part : We seek to find the temporal correlation between two companies. Therefore, we would like to combine the information from news, markets, and market segmentation to do prediction. Also, we want to take the characteristics of companies into consideration to reweigh the parameters.capture period(s) - feature selections are based on the papers (5),(6)Capture events’ affection - feature selections are based on the papers (5), (6) In the papers (1),(2), SVM applied in time series data was proposed / stated as a promising approach to predict financial market prices. Our hypothesis is that there exist some periods or event affection in the financial market prices, and we aim to extract them. Thus, we plan to write our own control algorithm to capture the periodic features with the implementation of SVR.4. References(1) Tay, F. E., & Cao, L. (2001). Application of support vector machines in financial time series forecasting. Omega, 29(4), 309-317.(2) Trafalis, T. B., & Ince, H. (2000). Support vector machine for regression and applications to financial forecasting. In Neural Networks, 2000. IJCNN 2000, Proceedings of the IEEE-INNS-ENNS International Joint Conference on (Vol. 6, pp. 348-353). IEEE.(3) Oldewurtel, F., Parisio, A., Jones, C. N., Gyalistras, D., Gwerder, M., Stauch, V., ... & Morari, M. (2012). Use of model predictive control and weather forecasts for energy efficient building climate control. Energy and Buildings, 45, 15-27.(4) Sapankevych, N. I., & Sankar, R. (2009). Time series prediction using support vector machines: a survey. IEEE Computational Intelligence Magazine, 4(2).(5) Kazem, A., Sharifi, E., Hussain, F. K., Saberi, M., & Hussain, O. K. (2013). Support vector regression with chaos-based firefly algorithm for stock market price forecasting. Applied soft computing, 13(2), 947-958.(6) Choudhury, S., Ghosh, S., Bhattacharya, A., Fernandes, K. J., & Tiwari, M. K. (2014). A real time clustering and SVM based price-volatility prediction for optimal trading strategy. Neurocomputing, 131, 419-426.(7) Ahmadi, P., & Samsami, F. (2010). Pharmaceutical market segmentation using GA K-means. European Journal of Economics, Finance and Administrative Sciences, 22, 72-83.(8) Hsu, S. H., Hsieh, J. P. A., Chih, T. C., & Hsu, K. C. (2009). A two-stage architecture for stock price forecasting by integrating self-organizing map and support vector regression. Expert Systems with Applications, 36(4), 7947-7951. (9) Xiaodong Li ? Haoran Xie ? Ran Wang(2016). Empirical analysis: stock market prediction via extreme learning machine ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download