Website is still under construction and missing some important links (April 19th, 2022)

This page provides information about the time series data mining studies during my PhD as well as the recent studies. This link summarizes the performance of the studies on the datasets from UCR time series database.

Recent work

1.Time series similarity based on a pattern-based representation (LPS)

A novel pattern-based time series representation along with a similarity measure is proposed in this study. We compare LPS with nearest neighbor classifiers with spatial assembly distance (SpADe), Dynamic Time Warping (NNDTW) distance. This work is submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) in 2013. After receiving a weak reject, the worked has been revised, improved and resubmitted to Data Mining and Knowledge Discovery and accepted in June of 2015. An R package, LPStimeSeries, is implemented as part of this study. This research was partially supported by the Scientific and Technological Research Council of Turkey (TUBITAK) grant number 114C103.

(link) (related files)

PhD work

1. A Bag-of-Features Framework to Classify Time Series (TSBF)

A feature based time series classification approach that can handle the translations and dilations in local patterns is proposed in this study. We compare TSBF with nearest neighbor classifiers with Dynamic Time Warping (NNDTW) distance. This work is published in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) in 2013.

(link) (related files)

2. Supervised Time Series Pattern Discovery through Local Importance (TS-PD)  

An exploratory approach to find the predictive regions of the time series for classification is proposed. TS-PD is compared to shapelet methods and NNDTW. This work received a major revision from Knowledge and Information Systems (KAIS) in 2013.

(link) (related files)

3- Multivariate Time Series Classification with Learned Discretization

A discretization strategy for multivariate time series (MTS) observations is introduced for time series classification.  Most of the studies working on MTS classification follow a different strategy for experimentation which makes the comparison of the approaches difficult. Therefore, we provide a database of MTS in the link given below to enable a fair comparison of the classification approaches. This work received a major revision from Data Mining and Knowledge Discovery (DAMI) in 2013 and accepted in 2014 after revisions. It is published in 2015.

(link) (related files)