Evolving possibilistic fuzzy modeling for financial interval time series forecasting

Financial interval time series (ITS) is a sequence of the highest and lowest values of financial data such as the highest and lowest prices of assets observed at successive time steps of a time interval. Price interval data carry key information to estimate price volatility, and provide valuable information to develop investment strategies. This paper suggests an evolving possibilisitc fuzzy modeling (ePFM) approach for financial ITS forecasting. ePFM employs memberships and typicalities to recursively cluster data, uses participatory learning to update the forecasting model structure as stream data is input, and processes interval-valued data. Computational experiments concerning the IBOVESPA index forecasting, the main equity market index of the Brazilian financial market, show that ePFM is a potential candidate for financial ITS forecasting. It achieves comparable or better performance than alternative interval-based approaches.