This study examines the performance of forecasting model established by using neural network. The neural network model proposed in this study is implemented to forecast the stock price stream of Korean stock market. Factors for forecasting that influence Korean stock market are obtained by the way based on multiple regression analysis. As a result of the analysis, Korean stock market is greatly influenced by the stock market of the USA. It is enough to be called co-movement phenomenon. In addition, the interest rate of Korea has a great effect on Korean stock market as well. So, we established the forecasting model of Korea Composite Stock Price Index (KOSPI) using neural network with Back-Propagation (BP) algorithm. Experimental results show that the proposed model has a good performance in forecasting KOSPI index.
According to globalization of the entire world, the economic situation of one country is influenced strongly by the economy of other countries. Korean stock market is receiving a lot of effects of oil price, interest rate of Korea and interest rate of the USA as well as the stock market of the USA, Japan, Europe etc.. Especially the stock market of the USA and Japan has a great effect on Korean stock market. This present state is more notable after foreign exchange crisis of Korea.
[...] The most popular sigmoid function for neural network is the hyperbolic tangent function. The tanh() function is a sigmoidal-shaped function and has the following symmetric shape. f x = Training of neural network model e x e x . It is essential to train a network for the success of neural network. Process that is converting initial weight value into the value which is suitable to data is proceeding in neural network model. Eq.(5) is used to calculate the error between output value calculated using activation function and target value. [...]
[...] These results indicate that the accurate rate of the proposed model is approximately Conclusion and future work The paper examines the factors that influences KOSPI index significantly and Evaluates neural network model that forecast KOSPI index using these factors. As a result of multiple regression analysis, KOSPI index is influenced significantly by call rate of Korea and NYSE composite index of the USA. R-square (adj.) calculated from this regression model indicates that explanation ability of the proposed regression model is fairly high. [...]
[...] Figure shows an undulating trend of the data used in multiple regression analysis 01/04/00 07/04/00 01/04/01 07/04/01 01/04/02 07/04/02 01/04/03 07/04/03 01/04/04 07/04/04 KOSPI index Fig KOSPI index 01/04/00 07/04/00 01/04/01 07/04/01 01/04/02 07/04/02 01/04/03 07/04/03 01/04/04 07/04/04 NYSE composite index Fig NYSE composite index daily interest rate of the USA 01/04/00 07/04/00 01/04/01 07/04/01 01/04/02 07/04/02 01/04/03 07/04/03 01/04/04 07/04/04 Fig daily interest rate of the USA daily interest rate of Korea 01/04/00 07/04/00 01/04/01 07/04/01 01/04/02 07/04/02 01/04/03 07/04/03 01/04/04 07/04/04 Fig daily interest rate of Korea We analyzed using multiple regression model whether these factors that influence Korean stock market are significant, Data are taken from daily value in the period of Jan until Dec in considering that Korean stock market has been influencing by various factors for changing of financial market circumstance from the progress of the opening of financial market after foreign exchange crisis of Korea. [...]
[...] Error term follows normal distribution Neural network model Neural network model is being used in the analysis of financial time series as they move from simple pattern recognition to a diverse range of application areas.[4] Neural network model is a information processing system that is composed of node, link, activity function and training algorithm etc . Usually, the structure of neural network model is built of input node, hidden node and output node. Input node receives data from the outside of system and transfers these data into system. [...]
[...] RMSE is defined as the value that expresses accurate rate of model and as following Eq.(7) 1 RMS = N N k=1 yk t k 2 These two criterions are used for evaluating the proposed model in this paper Results In forecasting time series data, a major problem is establishing the optimal number of nodes on each layer. The most common method in determining the number of hidden nodes is by trial-and-error or by means of experiments. So the optimal number of hidden nodes in this proposed model is selected by means of empirical testing. [...]
APA Style reference
For your bibliographyOnline reading
with our online readerContent validated
by our reading committee