The attribute reduction is one of the key processes for knowledge acquisition. Some data set is multidimensional and larger in size. If that data set is used for classification it may end with wrong results and it may also occupy more resources especially in terms of time. Most of the features present are redundant and inconsistent and affect the classification. In order to improve the efficiency of classification these redundancy and inconsistency features must be eliminated. This paper discusses an algorithm based on discernibility matrix and Information gain to reduce attributes.