Majority Voting Based Multi-Task Clustering of Air Quality Monitoring Network in Turkey

Abstract

Air pollution, which is the result of the urbanization brought by modern life, has a dramatic impact on the global scale as well as local and regional scales. Since air pollution has important effects on human health and other living things, the issue of air quality is of great importance all over the world. Accordingly, many studies based on classification, clustering and association rule mining applications for air pollution have been proposed in the field of data mining and machine learning to extract hidden knowledge from environmental parameters. One approach is to model a region in a way that cities having similar characteristics are determined and placed into the same clusters. Instead of using traditional clustering algorithms, a novel algorithm, named Majority Voting based Multi-Task Clustering (MV-MTC), is proposed and utilized to consider multiple air pollutants jointly. Experimental studies showed that the proposed method is superior to five well-known clustering algorithms: K-Means, Expectation Maximization, Canopy, Farthest First and Hierarchical clustering methods.

Wireless Access Control in Edge-Aided Disaster Response: A Deep Reinforcement Learning-Based Approach

“…In dynamic network scenario whose statistic information is hard to obtain in advance, the reinforcement learning technique [14] has been widely investigated to optimize its wireless resource allocation. Reinforcement learning technique is one of the three main paradigms in machine learning, besides supervised learning [15] and unsupervised learning [16]. In [17], Xu et al proposed an efficient reinforcement learningbased resource management algorithm, which learns on-thefly the optimal policy of dynamic workload offloading and autoscaling at network edge.…”

Evaluation of Multi-variate Time Series Clustering for Imputation of Air Pollution Data

“…Partitioning algorithms such as k-means and k-medoids are very common among works related to TS clustering and have been applied in many papers (e.g. Ignaccolo et al (2008); Austin et al (2013); Tuysuzoglu et al (2019)) 2 https://doi.org/10.5194/gi-2021-11 Preprint. Discussion started: 17 May 2021 c Author(s) 2021.…”

“…Ignaccolo et al (2008) transformed the TS of pollutant daily observations into a functional form to smooth the TS, then classified the air quality monitoring network in Northern Italy using the Partitioning Around Medoids algorithm (PAM) to cluster three individual pollutants, namely NO 2 , PM 10 , and O 3 . Tuysuzoglu et al (2019) applied different clustering algorithms such as k-means, Expectation Maximisation, and Canopy for each air pollutants in the dataset (NO, NO 2 , SO 2 , PM 10 , and O 3 ), then aggregated the clustering results based on majority voting to identify one clustering solution for similar regions in terms of air quality.…”

Evaluation of multivariate time series clustering for imputation of air pollution data

“…Partitioning algorithms such as kmeans and k-medoids are very common among works related to TS clustering and have been applied in many papers (e.g. Ignaccolo et al, 2008;Austin et al, 2013;Tuysuzoglu et al, 2019) Austin et al ( 2013) used the k-means algorithm to identify spatial patterns in air pollution data to cluster US cities based on the similarity of their PM 2.5 composition profiles, then characterize these clusters based on chemical characteristics, emission profiles, geographic locations, and population density. Ignaccolo et al (2008) transformed the TS of pollutant daily observations into a functional form to smooth the TS, then classified the air quality monitoring network in northern Italy using the partitioning around medoids algorithm (PAM) to cluster three individual pollutants, namely NO 2 , PM 10 , and O 3 .…”

“…Ignaccolo et al (2008) transformed the TS of pollutant daily observations into a functional form to smooth the TS, then classified the air quality monitoring network in northern Italy using the partitioning around medoids algorithm (PAM) to cluster three individual pollutants, namely NO 2 , PM 10 , and O 3 . Tuysuzoglu et al (2019) applied different clustering algorithms such as k-means, expectation maximization, and canopy for each air pollutant in the dataset (NO, NO 2 , SO 2 , PM 10 , and O 3 ), then aggregated the clustering results based on majority voting to identify one clustering solution for similar regions in terms of air quality.…”