Tuning Suitable Features Selection using Mixed Waste Classification Accuracy

Authors
Hassan Mehmood Khan1, Norrima Mokhtar1, *, Heshalini Rajagopal2, Anis Salwa Mohd Khairuddin1, Wan Amirul Bin Wan Mohd Mahiyidin1, Noraisyah Mohamed Shah1, Raveendran Paramesran3
1Department of Electrical Engineering, Faculty of Engineering, University of Malaya, Malaysia
2Department of Electrical and Electronic Engineering, Manipal International University, Malaysia
3Institute of Computer Science and Digital Innovation, UCSI University, 56000, Kuala Lumpur, Malaysia
*Corresponding author. Email: [email protected]
Corresponding Author
Norrima Mokhtar
Received 19 October 2020, Accepted 18 October 2021, Available Online 29 December 2021.
DOI
https://doi.org/10.2991/jrnal.k.211108.014How to use a DOI?
Keywords
Features reduction; features optimization; higher classification rate; mixed waste classification
Abstract
Classification accuracy can be used as method to tune suitable features. Some features can be mistakenly selected hence derailed the classification accuracy. Currently, feature optimization has gained many interests among researchers. Hence, this paper aims to demonstrate the effects of features reduction and optimization for higher classification results of mixed waste. The most relevant features with respect to mix waste characteristic were observed with respect to classification accuracy. There are four stages of features selection. The first stage, 40 features were selected with training accuracy 79.59%. Then, for second stage, better accuracy was obtained when redundant features were removed which accounted for 20 features with training accuracy of 81.42%. As for the third stage 17 features were maintained at 90.69% training accuracy. Finally, for the fourth stage, additional two more features were removed, however the classification accuracy was decreased to less than 80%. The experiments results showed that by observing the classification rate, certain features gave higher accuracy, while the others were redundant. Therefore, in this study, suitable features gave higher accuracy, on contrary, as the number of features increased, the accuracy rate were not necessarily higher.
Copyright
© 2021 The Authors. Published by ALife Robotics Corp. Ltd.
Open Access
This is an open access article distributed under the CC BY-NC 4.0 license (http://creativecommons.org/licenses/by-nc/4.0/).