An approach for Knowledge Acquisition from decision tables containing Continuous-Valued Attributes (KACVA) is developed.The equivalence partition in the classical rough sets theory is converted into a similarity partition.A novel representation method of the positive region in decision tables with continuous-valued attributes is built.Through calculating classification abilities of each conditional cluster to decision classes

decision rules in decision tables containing continuous-valued attributes are generated.Experimental evaluation on different data sets shows that the KACVA algorithm has the better performance in the classification accuracy comparing with the knowledge acquisition approaches under classical rough sets theory and the decision tree approach

C4.5

in processing decision tables with continuous-valued attributes.