Time series forecasting based on a neural network with weighted fuzzy membership functions

This paper proposes time series forecasting using a new feature selection method based on the non-overlap area distribution measurement method and Takagi´s and Sugeno´s fuzzy model. The non-overlap area distribution measurement method selects the minimum number of 4 input features with the highest performance result from 12 initial input features by removing the worst input features one by one. This paper proposes CPP_n,m (Current Price Position on day n: percentage of the difference between the price on day n and the moving average of the past m days´ prices from day n-1) as a new technical indicator. The performance result improves by from 58.35% to 58.86% when CPP_n,5 is added to the minimum number of 4 input features that are selected by the non-overlap area distribution measurement method as a new input feature.