Comparative analysis of spatio/spectro-temporal data modelling techniques

Abdullah, Mohd Hafizul Afifi and Othman, Muhaini and Kasim, Shahreen (2017) Comparative analysis of spatio/spectro-temporal data modelling techniques. In: Research Book – Data Engineering and Information Security Series 1. Penerbit UTHM, Batu Pahat, Johor, pp. 1-8. ISBN 9789672110583

Text Chapter 1_DEISS_S1.pdf Restricted to Registered users only Download (492kB) | Request a copy

Abstract

A fundamental challenge in spatio/spectro-temporal data (SSTD) is to learn the pattern and extract meaningful information that lies within the data. The close interrelationship between the space and temporal components of SSTD directly increases the complexity and challenges in modelling the data [1]. Other challenges include the dynamic pattern of spatial components features and inconsistency in the number of samples and feature-length used in the training and sampling datasets [2]. Data pre-processing method such as removal of irregular-feature data structure, however, may cause data loss which will lead to the final result become error prone. Despite the difficulties to process information from SSTD, several works on predictive modelling have been published, including applications on brain data processing [3], stroke data [4-5], forecasting of weather-driven damage in electrical distribution system [6], and ecological or environmental event prediction [7]. According to [8], environmental events often occur in a predictable temporal structure. Hence, the ability to exploit spiking neural network (SNN) by incorporating SSTD modelling techniques may be able to aid the process of discovering the hidden pattern and relationship between the two components of STTD; time and space. Recent work in [5], stated that most events occurring in nature form SSTD which requires measuring spatial or/and spectral components over time. Therefore, this paper presents the comparative analysis between various techniques used to process information from SSTD. Section 2 overviews two different inference-based techniques for SSTD modelling which includes global modelling, local modelling, and personalized modelling; and data modelling for SSTD classifier including, support vector machines (SVM), Evolving Classification Function (ECF), k-Nearest Neighbor (kNN), weighted k-Nearest Neighbor (wkNN), and weighted-weighted k-Nearest Neighbor (wwkNN). Section 3 presents the results of the assessment both SSTD inference-based modelling techniques and data training algorithms, while Section 4 concludes the analysis and ideas for future works.

Item Type:	Book Section
Uncontrolled Keywords:	Engineering—Management; Information technology—Security measures; Data protection
Subjects:	T Technology > T Technology (General) > T11.95-12.5 Industrial directories > T58.5-58.64 Information technology T Technology > TA Engineering (General). Civil engineering (General) > TA190-194 Management of engineering works
Divisions:	Faculty of Computer Science and Information Technology > Department of Software Engineering
Depositing User:	Mrs. Farizah Sudin
Date Deposited:	13 Jan 2022 07:24
Last Modified:	13 Jan 2022 07:24
URI:	http://eprints.uthm.edu.my/id/eprint/4345

Actions (login required)

View Item