基于自适应高斯混合模型的数据库基数估计方法OA

Cardinality estimation is a critical component of database query optimization,where its accuracy directly impacts the execution efficiency of query plans.Deep autoregressive model-based cardinality estimators have demonstrated remarkable accuracy in prior studies.However,they struggle to capture data distribution patterns when handling large-domain continuous attributes,which lead to significant performance degradation.To address these challenges,this paper proposed a novel cardi-nality estimator based on an adaptive Gaussian mixture model,called AGCard.It first dynamically adjusted the number and parameters of Gaussian components to adaptively fit the data distribution of continuous attributes,thereby reducing the domain scale.Subsequently,AGCard employed a bias correction algorithm to compensate for the estimation deviations introduced by the progressive sampling process while avoiding additional computational overhead.Extensive experiments on three real-world datasets(including WISDM)demonstrate that the proposed method outperforms existing mainstream baselines in terms of esti-mation accuracy,inference latency,and storage overhead.The results confirm the effectiveness of the adaptive Gaussian mix-ture model and the bias correction algorithm.