Estimating incomplete information in group decision making: A framework of granular computing