Multi-classifier models to improve accuracy of water quality application

• 主要な著者: Mokhairi, Makhtar, Mohd Nordin, Abdul Rahman, Mohd Khalid, Awang
• フォーマット: 論文
• 言語: English
• 出版事項: 2016
• 主題: QA75 Electronic computers. Computer science
• オンライン･アクセス: http://eprints.unisza.edu.my/7216/1/FH02-FIK-16-05680.jpg; http://eprints.unisza.edu.my/7216/

This paper presents a comparison among the different classifiers such as Naïve Bayes (NB), decision tree (J48), Sequential Minimal Optimization (SMO), Multi-Layer Perception (MLP), and Instance Based for K-Nearest neighbor (IBK) on water quality for datasets of Kinta River, Perak, Malaysia. Classification accuracy and confusion matrix were used in this research based on a 10-fold cross validation method. Then, a fusion at classification level between these classifiers was applied to get the highest accuracy and see which the most suitable multi-classifier approach for the datasets. The water quality datasets were taken from the East Coast Environmental Research Institute (ESERI) of University of Sultan Zainal Abidin (UniSZA). The water quality classes were evaluated using 10 factor indices, namely DO Sat, DO Mgl, BOD Mgl, COD Mgl, TS Mgl, DO Index, AN Index, SS Index, Class, and Degree of Pollution. The results showed that the classification using fusion between IBK+MLP, IBK+SMO, and IBK+MLP+NB+SMO was superior to the other classifiers that achieved the higher accuracy with the same percentage of 93.98%. Thus, using multiclassifier approaches can achieve better accuracy than the single ones.