| Author | Soontarin Nupap |
| Call Number | AIT Thesis no.IM-07-14 |
| Subject(s) | Stock exchanges--Thailand--Forecasting
|
| Note | A thesis submitted in partial fulfillment of the requirements for
the degree of Master of Science in Information Management, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. IM-07-14 |
| Abstract | Stock trading has been attracting a lot of people due to the high returns obtained by investing in stocks of large companies. Due to the sharp fluctuations in the price of the stock within a very short time, it is difficult to determine when to purchase or sell the stocks and which stock should be bought to achieve highest return. Several systems have been developed by researchers in order to effectively model the ever-changing behavior of the stock market. However, due to the inherent complexity in the stock pricing brought about by factors like the political unrest and environmental hazards, the prediction of the SET Index for a long term becomes unreliable. In this study, Genetic Algorithm (GA) and Neural Network (NN) have been integrated to improve the performance of NN. The stock forecasting system for supporting investors' decisions on the Stock Exchange of Thailand (SET) is proposed. Using Fundamental Analysis to analyze the inputs necessary to obtain the SET Index for a particular day, we propose to use the Interest Rate, Exchange Rate, Oil Price, Gold Price and the Date as the variables for computing the SET Index. From the analysis of the real data collected from the Stock Exchange of Thailand, the SET Index has been found to change drastically during a short period of time. In order to approximate the function to obtain the SET Index which uses the aforementioned five inputs as its variables, Neural Network has been used to train on the training set prepared from the real data. The inputs have been normalized and principal component analysis is performed on the inputs. The neural network has been trained with different number of neurons in the hidden layer to ensure that the result obtained is not a local minimum of the solution space. The neural network has been found to be highly effective in learning the weights and biases from the training set and good results have been obtained when testing with the test set containing real data. After obtaining the weights and biases, the necessary post processing is done to obtain the result in the required units. Genetic algorithm has been integrated into the system in order to obtain a good set of records with high quality rank and diversity rank. The output obtained by integration of Genetic Algorithm and Neural Network is better than just Neural Network alone but we need to ensure that the number of generations and crossover points is not very high. The stock forecasting system developed is able to forecast for one day using the previous day's inputs. The results obtained have high accuracy. The system also allows the user to specify the inputs for a particular date in order to compute the SET Index for that day by using stored weights and biases from the neural network or from the currently trained network |
| Year | 2007 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis ; no. IM-07-14 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Information Management (IM) |
| Chairperson(s) | Vatcharaporn Esichaikul; |
| Examination Committee(s) | Guha, Sumanta;Phan Minh Dung; |
| Scholarship Donor(s) | RTG Fellowship; |
| Degree | Thesis (M.Sc.) - Asian Institute of Technology, 2007 |