Materialized view is important to any data intensive system where answering queries at runtime is subject of interest. Users are not aware about the presence of materialized views in the system but the presence of these results in fast access to data and therefore optimized execution of queries. Many techniques have evolved over the period to construct materialized views. However the survey work reveals a few attempts to construct materialized views based on attribute similarity measure by statistical similarity function and thereafter applying the clustering techniques. In this paper we have proposed materialized view construction methodology at first by analyzing the attribute similarity based on Jaccard Index then clustering methodology is applied using similarity based weighted connected graph. Further the clusters are validated to check the correctness of the materialized views.