The concept and algorithm of the simplified discernment function are provided. The function not only has the same decision ability as the decision table

but also eliminates all the reduplicated and redundant terms included in the original discernment function educed by the decision table. To reduce the search space of the fitness function and improve the computing efficiency

an efficient genetic reduction algorithm is provided. By regarding the chromosome coverage of the simplified discernment function and the number of “1” in the chromosome as the parameters of the fitness function

it can be ensured that the algorithm will converge at the minimum reduction and improve the searching efficiency. It is proved theoretically that the attribute reduction calculated by the algorithm is optimal and the algorithm com

plexity is O(|f'||C||U|

2

). The algorithm is verified by four examples

and the results show that the number of terms of simplified discernment function is 0.39%

0.000 8%

0.000 08% and 0.000 3% of that of the original one respectively

and the minimal reductions can be obtained within 500 iterations.