Please use this identifier to cite or link to this item:
|Other Titles:||Forecasting Stock Index futures price using an application of quantum optimization with Fuzzy inference system|
|Publisher:||เชียงใหม่ : บัณฑิตวิทยาลัย มหาวิทยาลัยเชียงใหม่|
|Abstract:||The purpose of this study was (1) to compare the forecasting performance between the statistical model group represent by ARIMA model and the machine learning model group represent by ANN. SVM, S-ANFIS and WT-QPSO-ANFIS model in stock index futures forecasting in terms of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Standard Error of Mean (SEM), (2) to analyze the appropriate investment's proportion and risk of stock index futures portfolio. The data is daily stock index futures of 33 stock index futures covering 2964 trading days from January 1st, 2009 to May14th, 2020. Futures contracts of stock index futures had been trading remarkably worldwide in 2020. To mitigate the market risks and make the best of investments, predicting the movement of stock index futures is considered significant to ensure the potential gain in futures market trading. However, the accurate and precise prediction is not easy to obtain in the financial data which normally have high fluctuation. While the conventional econometric methods for forecasting require many assumptions, the machine learning models generally involve complex and unintelligible rules as well as a complicated network structure. In addition, the machine learning model itself did not guarantee a global optimum solution. It easily falls to the local optimum answer that directly affects the model's predicted value accuracy. To handle these drawbacks, in this study, we propose using a novel hybrid Wavelet Transformation-Quantum-behaved Particle Swarm Optimization-Adaptive Neuro-Fuzzy Inference System (WT-QPSO-ANFIS) model to forecast stock index futures. Generally speaking, three approaches, Wavelet Transformation (WT), Quantum Particle Swarm Optimization (QPSO), and Adaptive Neuro-Fuzzy Inference System (ANFIS), are combined in our forecasting model. The result reveals that the hybrid WT-QPSO-ANFIS model provides higher efficiency and accuracy in predicting 27 stock index futures considered in this study compared to the SVM model, ANN model, conventional Sugeno-type, ANFIS model, and ARIMA model consequently in terms of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Standard Error of Mean (SEM). Furthermore, risk analysis and portfolio management demonstrate that using the lowest variance approach based on stock index futures forecasting data from the most effective model provides a good rate of return than using the equal weighting method. Besides, the risk value calculated by using the Value at Risk and the Expected Shortfall is lower than the same equal weighting approach.|
|Appears in Collections:||ECON: Theses|
Files in This Item:
|611631011 นำโชค ฉิมปราง.pdf||15.19 MB||Adobe PDF||View/Open Request a copy|
Items in CMUIR are protected by copyright, with all rights reserved, unless otherwise indicated.