It is proved in the experiment that in this way superfluous knowledge can be removed easily symptoms can be reduced and a corresponding rule can be extracted .

u5b9eu9a8cu8bc1u660euff1au7528u6b64u65b9u6cd5u80fdu6709u6548u5730u53bbu9664 u5197u4f59u4fe1u606fuff0cu5bf9u5176u75c7u72b6u8fdbu884cu7ea6u7b80u63d0u70bcuff0cu4eceu800cu83b7u53d6u7b80u5355u800cu53c8u80fdu4f53u73b0u75c7u72b6u4e0eu75c5u5f81u7684 u89c4u5219u3002

Its purpose is to reduce superfluous attributes value to make the rule set more generalizable on the condition that it keeps classification capacity of the rule set .

u5b83u7684u76eeu7684u5c31u662fu5728u4fddu6301u89c4u5219u96c6u7684u5206u7c7bu80fdu529bu7684u6761u4ef6u4e0buff0c u5220u9664 u591au4f59u5c5eu6027u503cuff0cu8fdbu4e00u6b65u7b80u5316 u89c4u5219u96c6u3002