Robert Gwadera'S TALK ON DB SEMINAR

Title: Multi-Stream Join Answering for Mining Significant Cross-Stream Correlations

Abstract

Sliding-window multi-stream join (SWMJ) was proposed as a operation for correlating information from different streams. Although there has been an extensive research on SWMJ processing no research analyzed significance of SWMJ result in terms of correlation strength while such an analysis is necessary in practical applications including monitoring systems (e.g., video surveillance).

We provide the first significance analysis of the SWMJ result by focusing on the frequency of windows satisfying a given equijoin predicate as the most important parameter of the SWMJ result. We derived a formula for computing the expected frequency of windows satisfying a given equijoin predicate that can be evaluated in quadratic time in the window size given a proposed probabilistic model of the multi-stream. The formula can be used used twofold: (I) to assess the strength of correlation between the streams and (II) to answer ad hoc SWMJ queries given a probabilistic model of the streams.

In experiments conducted on a daily rainfall data set we demonstrate remarkable accuracy of our method, which confirms our theoretical analysis. 

 

Bio

Robert Gwadera is a senior researcher at Distributed Information Systems Laboratory, EPFL. He obtained Ph.D in Computer Science from Purdue University and before joining EPFL he was a Research Staff Member in Data Analytics at IBM Zurich Research Laboratory.