Rule extraction is performed on three kinds of time series. The first one is stock market data. The second one is artificially generated with a given covariance matrix using the Inverse Whitening Transformation. The covariance matrix is defined with a definite range of memory using the short memory form of exponential decay. The third one is also artificial generated while the covariance matrix is defined with long memory form of power decay. Classification of the time series into four categories is performed on these real value sequences of finite number of classes. The sequences are then divided into three subsets: training, testing and news set. Rules are extracted using genetic algorithm for forecasting time series in the training set. A test of the correct ratio for the rule is don...[ Read more ]

Rule extraction is performed on three kinds of time series. The first one is stock market data. The second one is artificially generated with a given covariance matrix using the Inverse Whitening Transformation. The covariance matrix is defined with a definite range of memory using the short memory form of exponential decay. The third one is also artificial generated while the covariance matrix is defined with long memory form of power decay. Classification of the time series into four categories is performed on these real value sequences of finite number of classes. The sequences are then divided into three subsets: training, testing and news set. Rules are extracted using genetic algorithm for forecasting time series in the training set. A test of the correct ratio for the rule is done in the testing set. Portfolio management is performed on the news set time series. We introduce two parameters to characterize different human nature of the best agent. These two parameters are the greediness and level of follow-up news. A universal behavior suggests that greedy and persistent investors perform better in portfolio management. We also introduce fuzzy behavior to the rules and similar patterns are found in portfolio management with the effect of greed and follow-up news.