Materialized view creation is an important aspect for large data centric applications. Materialized views create an abstraction over the actual database tables to the users. Users are not aware about the existence of these materialized views. However, these help in faster execution of query. Materialized views should contain the data that users are currently accessing, and possibly those that would be accessed in near future. Availability of the user-requested data in a materialized view indicates the efficacy of the materialized view creation process. A review of the existing research work reveals a gap in analyzing the inter-attribute affinity while creating the materialized views. This paper proposes a new methodology for materialized view creation by quantifying the association among the independent data attributes. This is done based on the usage of different attributes in the recently executed set of queries. Statistical analysis on existing query set help to predict the attributes likely to be used for future queries. The materialized views are generated accordingly. © 2012 IEEE.