Abstract: Because the conventional K-NN algorithm has the disadvantages that the result is easily disturbed by single attribute of record and the time efficiency is very low，a novel K-NN algorithm is proposed which is an enhanced K-NN algorithm using information gain and extension relativity.Get the role coefficient of attribute，by which the attributes can be classified into primary attribute，secondary attribute and redundant attribute，through computing the information gain of attribute.The role coefficient is introduced to the Euler’s distance equation so that the effect of every attribute is controled by weightiness of it，by that，the anti-jamming ability and accuracy of the K-NN algorithm are improved highly.Partition the attribute space into several sub-spaces，and map the record，which will be recognized，to a sub-space which a searching space is comprised of.So，the computing time is decreased and the time efficiency is improved greatly.The test results show the novel K-NN algorithm is feasible and effective.