Prediction of foreign exchange rate by local fuzzy reconstruction method

Several systems for the purpose of predicting trends in the foreign exchange market and stocks have been developed. They are either knowledge based expert systems or fuzzy expert systems. The disadvantage of these systems is that it mainly depends on knowledge base. It is not easy to obtain the know-how of dealers completely. This paper presents the method of predicting time series data, which is considered as a deterministic chaos, based on the deterministic dynamical system. The key technology is embedding and local reconstruction. At first, we explain about deterministic chaos. Next, we show the Takens' embedding theorem briefly, and discuss the local reconstruction especially the local fuzzy reconstruction method in detail. Finally. we apply the local fuzzy reconstruction method to the prediction of foreign exchange rate, and the results obtained are included.