Indexed by:会议论文

Date of Publication:2011-01-01

Included Journals:CPCI-S

Page Number:23-+

Key Words:Data stream; Synopsis data; Random sampling; Sliding window; DWRS algorithm

Abstract:In order to meet the requirement of synopsis data structure to process massive data, this paper presents a dynamic weighted random sampling algorithm (DWRS) on time-based sliding window on the improvement of weighted random sampling algorithm over data stream. To reduce the influence of data stream instability on the accuracy of synopsis data, we calculated jitter factor and characteristic value dynamically when data stream arrives, and then calculated weight of every data item according to jitter factor and characteristic value. After that we constructed synopsis data by sampling method over weighted data stream. Experimental results show that the novel algorithm is efficient for the construction of synopsis data structure over data stream.