Self-adaptive sampling and querying method and system (Chinese)

The invention provides a self-adaptive sampling and querying method and system, and the method comprises the steps: calculating the fluctuation weight of each data point based on a cosine distance weight function of fluctuation, and carrying out the quick self-adaptive bucket separation of accumulated data by adopting a binary search algorithm according to the fluctuation weight, so that the sum of the maximum weights of all buckets in the bucket dividing result is minimum; for streaming data in each bucket dividing result, extracting samples with the same number through a plurality of sampling operators so as to realize adaptive sampling density and obtain a corresponding sampling result; based on a sampling result, sampling from the low-level samples to obtain high-level samples, constructing a hierarchical sample structure which keeps hierarchical result consistency, and constructing a hierarchical query engine according to the hierarchical sample structure; and after the hierarchical query engine queries the hierarchy conforming to the sampling granularity, keeping the sample of the current hierarchy as a query result. According to the invention, the sampling error is reduced, and the consistency and low delay of the data query result are ensured.

本发明提供一种自适应采样与查询方法及系统，该方法包括：基于波动度的余弦距离权重函数，计算每一个数据点的波动度权重，并按照所述波动度权重，使用二分查找算法对累积数据进行快速自适应分桶，使得分桶结果中所有桶的最大权重和最小；对每个分桶结果中的流式数据，通过多种采样算子提取相同数目的样本，以实现自适应的采样密度，得到对应的采样结果；基于采样结果，从低层级样本中采样得到高层级的样本，构建保持层级结果一致性的层级样本结构，并根据所述层级样本结构，构建层级查询引擎；当层级查询引擎查询到符合采样粒度的层级后，将保留当前层级的样本作为查询结果。本发明降低了采样误差，保证了数据查询结果的一致性与低延迟。