E-Locus - Institutional Repository of the University of Crete

Home Search

Results - Details

Search command : Author="Τσακαλίδης" And Author="Παναγιώτης"

Current Record: 17 of 66

[Add to Basket]

Identifier

000430858

Title

Data-driven symbolic representations for high-level time series analysis

Alternative Title

Συμβολικές αναπαραστάσεις βάσει δεδομένων για ανάλυση χρονοσειρών σε υψηλό επίπεδο

Author

Μπουντρογιάννης, Κωνσταντίνος Ε.

Thesis advisor

Τσακαλίδης, Παναγιώτης

Reviewer

Τζαγκαράκης, Γεώργος
Καρυστινός, Γεώργιος
Τζίτζικας, Ιωάννης

Abstract

The systematic collection of data has become an intrinsic process of all aspects in modern life. From industrial to healthcare machines and wearable sensors, an unprecedented amount of data is becoming available for mining and information retrieval. The ever-increasing volume and complexity of time series data necessitate efficient dimensionality reduction for facilitating data mining tasks. Symbolic representations, especially the family of symbolic aggregate approximations (SAX), have proven very effective in compacting the information content of time series while exploiting the wealth of search algorithms used in bioinformatics and text mining communities. However, typical SAX-based techniques rely on a Gaussian assumption for the underlying data statistics, which often deteriorates their performance in practical scenarios. To overcome this limitation, this thesis introduces a method that negates any assumption on the probability distribution of time series, by means of kernel density estimation (KDE) and Lloyd-Max quantization. Experimental evaluation on real-world datasets demonstrates the superiority of the proposed method, when compared against the conventional SAX and an alternative data-adaptive SAX-based method. Finally, in the present thesis, the proposed dimensionality reduction method is utilized to provide compact representations of time series for the purposes of anomaly detection. To this end, a computationally efficient, yet highly accurate, framework for anomaly detection of streaming data in lower-dimensional spaces is developed, whereas alternative quantization schemes are explored and utilized for more accurate statistical inference.

Language

English

Subject

Data-driven probabilistic SAX

Kernel density estimation

LLOYD-MAX quantization

Mode-bounding quantizer

Online anomaly detection

Symbolic representations

Ανίχνευση ανωμαλιών σε πραγματικό χρόνο

Εκτίμηση πυκνότητας πιθανότητας

Κβάντιση LLOYD-MAX