This paperanalyzed the information asymmetry problem between prediction areas and known areas faced by the mineral resource quantitative predictionand brought a comprehensive geo information quantitative prediction method limited by spatial affecting extent to deal with it, which applied in manganese ore prediction in western Guangxi and southeastern Yunnan, China.Its major steps included: (1) The spatial affecting extent of geo variant was defined with intersection of data extent, affecting extent and extent occupied by range of values of known areas, avoiding extrapolationof raw data when not occupy whole prediction areas; (2) The spatial affecting extent of prediction model was defined with intersection of geo variants that that make up the model, and model was meaninglesswhen the intersectionis null; (3) Through combinationof geo variants, a series of quantitative prediction models was built using multivariate linear regression; (4) To each prediction unit, prediction model was selected to calculateits reserve, which including most geo information and best regression effect. The procedure of prediction matched metallogenesisof prediction areas with prediction models built in known areas, which efficientlyensured the information symmetry.