There have been a series of debates to determine whether it would be possible to forecast dynamic systems such as stock markets. Recently the introduction of chaos theory has allowed many researchers to bring back this issue. Their main concern was whether the behavior of stock markets is chaotic or not. These studies, however, present divergent opinions on this question, depending upon the method applied and the data used. And the issue of predictability based on the nonlinear, chaotic nature was not dealt extensively. This paper is to test the nonlinear nature of the Korea stock market and accordingly attempts to predict its behavior. The result indicates that our stock market represents a chaotic behavior. We also found out based on our simulation that executing buy/sell transactions based upon forecasts which were derived using the local approximation method outperforms the decision of holding without a buy/sell transaction.