DS Ph.D. Qualifier Presentation | Xiaoshuai (Maksim) Li | Thursday, April 24th @ 8:00am via zoom

DATA SCIENCE 

Ph.D. Qualifier Presentation

Xiaoshuai (Maksim) Li

When: April 24th Wed: 8:00 AM-9:00 AM

Virtual Meeting: contact datascience@wpi.edu for the zoom link

Committee:

Prof. Mohamed Eltabakh (advisor)

Prof. Elke Rundensteiner (co-advisor)

Prof. Oren Mangoubi

Title: Regular Expression-Based Similarity Queries over Big Time Series Data 

Abstract:

Time series data is pivotal across various domains, including finance, healthcare, sensor networks, and IoT, underscoring the importance of analyzing and understanding trends and patterns in time series. Similarity search, fundamental to analytical tasks such as clustering, classification, forecasting, and pattern detection, typically relies on users specifying exact sequences of values. This approach imposes significant limitations: firstly, users must know precisely what values to search for; secondly, it restricts them to specific rather than more flexible and powerful matching patterns. This paper addresses these constraints by introducing a novel and powerful Regular-Expression-based query search language tailored assisting users expressing requirements of query patterns, accommodating the unique needs of time series applications. Further, we develop both a novel dual-layered indexing system and query processing optimized for efficiently processing these queries across large-scale distributed time series datasets. Our experiment over the Random Work datasets shows that our proposed approach significantly reduces overall query processing time and improves query performance.