LAQP: Learning-based approximate query processing

The querying on big data is a challenging task due to the rapid growth of amount. Approximate query processing (AQP) way meet requirement fast response. In this paper, we propose learning-based AQP method called LAQP. LAQP builds an error model learned from historical queries predict sampling-based estimation each new query. It makes combination AQP, pre-computed aggregations and provide high-accurate estimations with small off-line sample. experimental results demonstrated that our outperforms pre-aggregation-based most recent method.