Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Best of 2016 Recommended by Editor Recommended by Reviewer Recommended by Reader
On searching and indexing sequences of temporal intervals
Kostakis O., Papapetrou P.  Data Mining and Knowledge Discovery 31 (3): 809-850, 2017. Type: Article
Date Reviewed: Oct 27 2017

Have you ever wondered how it could be possible for a robot and its sensory system to understand obstacles and avoid them while randomly moving around? Did you ever ask yourself the questions of how signals can be captured and interpreted in a meaningful way, or whether a computer really understands the emergence of events? If you imagine that events are conceived as a series of time intervals, for instance, on a signal, probably from outer space, then this paper could be a good starting point to make up your mind on what the underlying technology and algorithmic approach for dealing, computationally, with time intervals may look like.

The paper could be considered as a contribution to the field of pattern recognition and image processing techniques, in which several approaches have been exploited in various application domains, including communication, medicine, data mining, audio and visual applications, sensors, and so on. These pattern recognition algorithms aim at achieving a reasonable match among the given input and the varying collection of data. Particularly, pattern recognition over time series data has been a research cornerstone for many years, and it will likely continue to be. How this is accomplished is by learning to distinguish patterns of interest from their classified groups, and then extracting the items reflecting the input pattern as accurately as possible.

Regardless of the advancements made in this field, there is still room for improvement due to the lack of a formal solution to effectively describe the features/patterns. Furthermore, the distance measuring tools for pattern recognition such as k-nearest neighbors or radial basis function typically rely on simple distances that do not provide meaningful results on the recognized patterns. The main research question is how to develop a pattern recognition solution capable of processing and clustering patterns, particularly among large and complex time series data, because this process turns out to be a computationally time-consuming task.

Within this context, this paper provides some interesting insights into the way patterns of time intervals can be indexed in order to be efficiently recognized. From the perspective of scholars, the paper provides an excellent entry point into the field of event-specific pattern search and indexing with events being conceived as sequences of time intervals. From a researcher’s point of view, the paper provides an excellent example of how to conduct research based on quantitative methods because it relies on experiments and carefully selected datasets. Hence, it considerably alleviates the task of validating the results or even reproducing the experiments and results, something that has become a rarity in scientific publications of low quality these days. Nonetheless, the paper has significant theoretical merits in the way algorithms and their complexity are discussed.

In summary, the paper definitely addresses a significant problem; however, it should be considered as incremental rather than groundbreaking research. Marginal improvements in the performance and effectiveness of retrieving time intervals without the deeper understanding of the event type will remain constrained in their impact on very important application domains such as sign language, robot sensors, sensor networks, and medicine.

Therefore, the main beneficiaries of the information in this paper are researchers and practitioners in academia or industry focusing on automated systems, where recognition and understanding of events is of major concern. Furthermore, academics and students involved in postgraduate studies with specialization in machine learning, data science, or big data analytics may benefit from reading this paper as well.

Reviewer:  Epaminondas Kapetanios Review #: CR145624 (1801-0018)
Bookmark and Share
  Editor Recommended
Data Mining (H.2.8 ... )
Information Storage And Retrieval (H.3 )
Would you recommend this review?
Other reviews under "Data Mining": Date
Online estimation of discrete, continuous, and conditional joint densities using classifier chains
Geilke M., Karwath A., Frank E., Kramer S.  Data Mining and Knowledge Discovery 32(3): 561-603, 2018. Type: Article
Dec 6 2018
Mining urban events from the tweet stream through a probabilistic mixture model
Capdevila J., Cerquides J., Torres J.  Data Mining and Knowledge Discovery 32(3): 764-786, 2018. Type: Article
Jul 5 2018
User preference learning for online social recommendation
Zhao Z., Lu H., Cai D., He X., Zhuang Y.  IEEE Transactions on Knowledge and Data Engineering 28(9): 2522-2534, 2016. Type: Article
Apr 23 2018

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright © 2000-2018 ThinkLoud, Inc.
Terms of Use
| Privacy Policy