DBMS Course Project Specifications

(under Prof. Krithi Ramamritham)

Title : Design and implementation of a method for finding aggregates using a single pass in streaming data

Abstract : For streaming data finding aggregates in single pass over the data without any access to past data is of crucial importance. This is particularly true for the data streams encountered in the telecom industry, stock exchange etc. The method of using the Wavelet Transforms to obtain a "sketch" of the data stream gives better approximations than the other methods of using the Histogram and sampling techniques. The wavelet transform methos consists of obtaining a specific linear projection of underlying data. It uses small amounts of summarization space and per item time while streaming through the data and provides accurate representation.

Deliverables : Implementation of the Wavelet tranform based method for finding aggregates for simple point, range and aged queries using a single pass over streaming data.

Presentation Slides PPT PDF (Customized for presentation I gave later elsewhere)

Experimental Results for Different Threshold Values

References :
Surfing Wavelets on Streams: One-Pass Summaries for Approximate Aggregate Queries
Wavelet-Based Histograms for Selectivity Estimation