In past decade, there is a huge growth in the field of technology. Attribute reduction for big data is viewed as an important preprocessing step in the areas of machine learning, Information mining and pattern recognition. MapReduce is useful programming model for data-intensive distributed parallel computing. In this paper, a K-mean clustering based MapReduce is proposed. This approach will help to do classification more accurately. The authors' model with MapReduce operations allows them to parallelize large computations effortlessly.