Online scalable SVM ensemble learning method (OSSELM) for spatio-temporal air pollution analysis

Loading...
Thumbnail Image
Other Title
Authors
Ali, Shahid
Dacey, Simon
Author ORCID Profiles (clickable)
Degree
Grantor
Date
2017
Supervisors
Type
Journal Article
Ngā Upoko Tukutuku (Māori subject headings)
Keyword
Auckland (N.Z.)
support vector machine (SVM)
SVM
Online Scalable SVM Ensemble Learning Method (OSSELM)
ensemble learning
air pollution analysis
spatio-temporal
aggregation (machine learning)
scalable
New Zealand
ANZSRC Field of Research Code (2020)
Citation
Ali, S., & Dacey, S. (2017). Online Scalable SVM Ensemble Learning Method (OSSELM) for Spatio-Temporal Air Pollution Analysis. International Journal of Data Mining & Knowledge Management Process, 7, 21-38. doi:10.5121/ijdkp.2017.7602
Abstract
Environmental air pollution studies fail to consider the fact that air pollution is a spatio-temporal problem. The volume and complexity of the data have created the need to explore various machine learning models, however, those models have advantages and disadvantages when applied to regional air pollution analysis, furthermore, most environmental problems are global distribution problems. This research addressed spatio-temporal problem using decentralized computational technique named Online Scalable SVM Ensemble Learning Method (OSSELM). Evaluation criteria for computational air pollution analysis includes: accuracy, real time & prediction, spatio-temporal and decentralised analysis, we assert that these criteria can be improved using the proposed OSSELM. Special consideration is given to distributed ensemble to resolve spatio-temporal data collection problem (i.e. the data collected from multiple monitoring stations dispersed over a geographical location). Moreover, the experimental results demonstrated that the proposed OSSELM produced impressive results compare to SVM ensemble for air pollution analysis in Auckland region.
Publisher
AIRCC Publishing Corporation
Link to ePress publication
DOI
doi:10.5121/ijdkp.2017.7602
Copyright holder
Authors
Copyright notice
All rights reserved
Copyright license
This item appears in: