Title: A proposed model based on k-nearest neighbour classifier with feature selection techniques to control and forecast plant disease
Authors: Inas Ismael Imran; Rawaa Hamza Ali; Shymaa Mohammed Jameel; Refed Adnan Jaleel
Addresses: Department of Computer, College of Education for Women, University of Baghdad, Baghdad, Iraq ' Department of Biology, College of Science, University of Misan, Amarah, Maysan, Iraq ' Iraqi Commission for Computers and Informatics, Baghdad, Iraq ' Info. and Comm. Engineering, Al-Nahrain University, Baghdad, Iraq; Commission of Care for Persons with Disabilities and Special Needs, Ministry of Labour and Social Affairs, Baghdad, Iraq
Abstract: Plant diseases have caused destruction. Main issues of today's agricultural managers is implementing effective discoveries that account for health of crops. Agricultural management starts to make sure plant gets right care at right time. This study used data from soya beans in order to create a classification model for forecasting plant status using the k-Nearest Neighbours (K-NN). Effectiveness of classifier drops when data contains noisy characteristics. The right attributes chosen lead to better predictions. Thus, the aim is to obtain a high-quality model and identify features that have greatest impact on status. The first experiment was carried out without use of feature selection and the second experiment was run with features selection techniques that are implemented in WEKA to assist in making decisions to forecast and improve plant health. Lastly, we evaluate a suggested framework in two experiments using k-fold cross-validation. K-NN combined with ReliefFAttributeEval performs exceptionally well.
Keywords: soya bean data; K-NN; feature selection; classification; evaluating.
DOI: 10.1504/IJGUC.2024.140109
International Journal of Grid and Utility Computing, 2024 Vol.15 No.3/4, pp.306 - 313
Received: 02 May 2023
Accepted: 08 Aug 2023
Published online: 24 Jul 2024 *