Detecting clusters or communities in large real-world graphs such as large social or information networks is a problem of much interest. The popularity of large on-line networks presents a new challenge to apply structural data analysis techniques in order to extract useful information. Dimensionality reduction can be used to improve both efficiency and effectiveness while extracting information from data. In this paper, we propose an algorithm to reduce the dimensionality of datasets such that after applying data mining techniques on reduced datasets, we get almost same results as with the original datasets. Random sketch is used to reduce the dimensions of the dataset. For evaluation of the proposed approaches on complex network, we apply it in community detection problem. The results obtained using our methods are very competitive with various existing well known algorithms. This is verified on a collection of real networks. On the other hand, it can be observed that time taken by our algorithm is much less compared to available popular methods.