Learning-Based Processing Of Top-N Queries On Data Streams

The study of data stream, we know that data stream with real-time, continuity, universality, semantic uncertainty and other features. This paper summarizes the advantages and disadvantages of the traditional data stream processing techniques, such as: histogram method, sampling method, the hash method and Wavelet method. Based on these method and according to the characteristics of data stream, this paper proposes the use of based on sliding window model constructed the summary database, this method overcome the limitations of the traditional technique dealing with data stream problem. Make use TOP-N query based on learning solve the data stream problems provide possible. Then, the paper summarizes the advantages and disadvantages of the traditional TOP-N query. Based on these method , the paper proposes a TOP-N query based on learning. The method first need to construct a knowledge database to store the query profile, after the Knowledge database Established, search the knowledge database directly. Knowledge database in the search, you must calculate the densityρ, then can get the query radius r by the densityρ,thus we can get the requirements of the N results approximate. When a new batch of data into the summary database, we needed to use some strategy update the knowledge database and the summary database.