Automatic Document Classification (ADC) becomes an increasingly important tool for helping people to organize the vast amount of data. ADC has been used in a number of different applications such as automatic indexing, content management and search pace categorization. In this paper, the authors introduce an efficient unsupervised learning approach which can cluster words into best significant similar groups. It will determine with the help of minimum overlap of the clusters. This paper also tries to resolve the problem of conventional classifiers which faces trouble to classify the high dimensional datasets and provide efficient accuracy through unsupervised learning method based classification.