Indexing is one of the most effective techniques for relational databases to achieve fast queryprocessing. The intelligent index tuning technique can effectively adjust the index of the database instance to obtain efficient query performance. Most of the existing methods utilize the query log to generate candidate indices, and then use the artificially designed models to select indices, thereby the indices are adjusted. However, the candidate indices generated from the query log may not exist in the database instance, so they cannot precisely estimate the effects of such indices on the query processing. This study first designs and implements an intelligent index tuning system for the relational database. Secondly, it proposes a learning-based method to model the effects of indices for query processing, accordingly, the query optimization effect of an index can be accurately estimated when selecting optimized indices. Then, an efficient optimal index selection algorithm is designed to select a set of indices with the maximal utility from candidate indices, which satisfy the space threshold. Finally, experiments are conducted to test the performance of the proposed system in different settings. The experimental results show that the proposed technique can effectively adjust the index and achieve a significant improvement in query performance for a relational database.