Study on Stock Market Prediction by Using Artificial Neural Network on High-Performance Computer
碩士 === 中華科技大學 === 機電光工程研究所碩士班 === 101 === Recently, the artificial neural network (ANN) is applied in different fields including, technology and financial engineering. Artificial neural network is easy to describe to nonlinear systems with having parallel processing furthermore, the system is easy t...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2013
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/87095966804170643951 |
| Summary: | 碩士 === 中華科技大學 === 機電光工程研究所碩士班 === 101 === Recently, the artificial neural network (ANN) is applied in different fields including, technology and financial engineering. Artificial neural network is easy to describe to nonlinear systems with having parallel processing furthermore, the system is easy turning.
The predication of TAIEX is investigated by using artificial neural network with some technical indicators. The index of stock market is affected by some variables and therefore the neural element is selected from by using regression analysis model and as the new neurons in the method of back-propagation neural network. The code is developed in MATLAB to achieve the calculation of ANN. The more data is leaned in artificial neural network with having higher accuracy probability. We turn the MATLAB code into C and the code is parallelized with OpenMP developed on modern high performance computing facilities platform. Based upon the results, the hitting rate is near 80% with considering 19 variables as input and 10 internal nodes in ANN scheme. Due to fewer nodes are used in this work, the performance is insignificant. The parallel efficiency is improved until the number of node is above 51 and the efficiency is near 50%.
|
|---|