Abstract: ROUSTIDA has highly ability of packing missing data.But it still has some incomplete information.This paper takes advantage of the discrimination of attributes suggested from extended discriminable matrix.The improved ROUSTIDA extends the ability of packing missing data，and has a new ability to eliminate noise data．It also reduces running time.All that has been proved in experiments.

摘要： ROUSTIDA算法具有较好的数据填补能力，但依然会出现一些不完备信息。利用了可扩充辨识所反映的对象间的属性差异信息，对遗失属性进行填充，从而使改进后的ROUSTIDA算法的填充能力得到了很大的改善，同时还具备了初步排除噪声数据的能力，在性能上也有了很大的提高，实验表明改进的算法具有很好的实用价值。