An Effective Approach for Processing Data Streams Queries

Abstract

Most of recent applications such as sensor networks generate continuous and time varying data which are called data streams. Additional constraints are faced for efficient query processing of such data streams that have uncertain nature and require fast and timely processing. Traditional query processing techniques of static data process the whole data without partitioning them, which is not applicable to data streams. Applying data clustering is demanded as a preprocessing step of data streams. Also, data streams are often suffer from incompleteness and high dimensionality. So, in this thesis, we introduce a framework for efficiently answering incomplete high dimensional data streams queries.

The proposed framework handles the incompleteness issue by estimating missed values based on the corresponding nearest-neighbors’ intervals. The continuous clustering mechanism is adopted and extended to accurately handle the incomplete data streams. The performance of incomplete data clustering was improved by 20% compared to the alternative approaches using two different data sets.

The proposed framework provides an improved subspace clustering to deal with high dimensional data streams. The experimental results using two datasets proved the efficiency of the proposed framework on average by 7.9% over the comparing algorithms for clustering such incomplete high dimensional data streams. The query processing performance of the incomplete high dimensional data streams was improved by 62% using two different data sets over the compared algorithms due the proposed clustering improvements of such data.