A particle swarm optimization based ensemble for vegetable crop disease recognition
Introduction
Machine learning algorithms enable effective decision making (Wolpert, 1992, Alpaydin, 2004, Chaudhary et al., 2016a, Chaudhary et al., 2016b, Uddin et al., 2019) when used for the cases of high dimensional agriculture data (Chaudhary et al., 2013a, Chaudhary et al., 2013b, Liakos et al., 2018, Rangarajan et al., 2018, Lawrence et al., 2020). The algorithms efficiently mine the complex relationships in the data (Rocha et al., 2010). The feature selection methods help in choosing the most relevant features from the big datasets (Timmermans and Hulzebosch, 1996, Kundu et al., 2011, Hill et al., 2014, EI-Bendary et al., 2015). Researchers showed that Logistic Regression and Naïve Bayes correctly identify the plant diseases (Baker and Kirk, 2007, Gutiérrez et al., 2008, Sankaran et al., 2010, Phadikar et al., 2013).
Brinjal, Beet, Cabbage, Celery, Chilli, French bean, Okra, Onion, Turnip, Potato, Tomato and Pepper are the vital vegetable crops. An important reason for their unstable and less production is the incidence of pest infections and diseases. Different bacteria, fungi, viruses, nematodes and physiological disorders are responsible for diseases. Exact recognition of disease is a multiclass classification problem.
Present work is conducted for classification of diseases using data samples for Anthracnose, Bacterial wilt, Black leg, Black-rot, Chilli-mosaic, Club-root, Downy-mildew, Early blight, Fusarium wilt, Gray mold, Late blight, Leaf-spot, Onion-smut, Powdery-mildew, Rust, Septoria leaf spot, Verticillium wilt, Yellow vein mosaic.
The ensembles classify better than the individual machine learning algorithms (Hansen and Salamon, 1990, Schapire, 1990, Breiman, 1996, Ho, 1998, Bay, 1999, Opitz, 1999, Ting and Witten, 1999, Zheng and Webb, 1999, Dietterich, 2000, Stamatatos and Widmer, 2005, Kotsiantis, 2007, Sun et al., 2007, Bolón-Canedo et al., 2012, Hsu, 2012, Farid et al., 2014).
The present work suggests a new EnsPSO approach with intent to enhance the performance outcomes of Vote. The EnsPSO is a combination of (i) Vote, (ii) CFS method, (iii) PSO algorithm and (iv) random sampling method. The work also presents performance comparison of newly proposed EnsPSO approach with Vote ensemble. The EnsPSO approach is applied for recognition of vegetable crop diseases. Section 2 describes the details of the materials and methods used. Section 3 describes proposed EnsPSO approach. Section 4 presents the results and discusses them. Section 5 summarizes the conclusions drawn.
Section snippets
Materials and methods
Present work is conducted using WEKA (Witten and Frank, 2005, Hall et al., 2009). WEKA consists of various supervised and unsupervised machine learning algorithms. It provides an extensive set of data pre-processing and modeling methods.
The proposed EnsPSO approach
The EnsPSO is intended to improve the disease classification accuracy as compared to Vote. The pseudo- code of EnsPSO is shown in Algorithm 1. Consider a disease dataset as D = Dtraining ∪ Dtesting where D contains disease influencing features, and the resultant diseases.Algorithm 1. EnsPSOInput: Dtraining = {S, F, C} where S = {s1, s2,…,sn} is non empty finite set of n samples such that for each sk ∈ S, 1 ≤ j ≤ n, has m features F = {f1, f2,…, fm} such that for each fp ∈ F, 1 ≤ p ≤ m, and
Results and discussion
Consistent estimates for classification accuracy are obtained using 10-fold cross validation strategy for assessing the performance of any machine learning algorithm (Baldi et al., 2000, Azar et al., 2014). The 10-fold cross validation strategy in present work divides the vegetable crop disease dataset into 10 pieces or 10 folds. The strategy used in the present work results in 10 evaluation results which are then averaged. Hence, each experiment is performed using 10-fold cross validation
Conclusions
The EnsPSO approach is an important contribution of present work. The EnsPSO presented for multiclass classification problems successfully recognize the vegetable crop diseases. The EnsPSO scales up the disease classification accuracy to 96% as compared to Vote which yields 84% accuracy. The EnsPSO shows better performance as compared to Vote for other performance measures as well. The EnsPSO approach is also tested for classification accuracy with 10-fold cross validation strategy on 3
CRediT authorship contribution statement
Archana Chaudhary: Conceptualization, Methodology, Data curation, Writing - original draft. Ramesh Thakur: Visualization, Investigation, Formal analysis. Savita Kolhe: Validation. Raj Kamal: Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (60)
- et al.
A random forest classifier for lymph diseases
Comput. Methods Programs Biomed.
(2014) - et al.
Comparative analysis of models integrating synoptic forecast data into potato late blight risk estimate systems
Comput. Electron. Agric.
(2007) - et al.
Democracy in neural nets: voting schemes for classification
Neural Networks
(1994) Nearest neighbor classification from multiple feature subsets
Intell. Data Anal.
(1999)- et al.
An ensemble of filters and classifiers for microarray data classification
Pattern Recogn.
(2012) - et al.
A hybrid ensemble for classification in multiclass datasets: an application to oilseed disease dataset
Comput. Electron. Agric.
(2016) - et al.
An improved random forest classifier for multi-class classification
Inform. Process. Agric.
(2016) - et al.
Hybrid decision tree and naive Bayes classifiers for multiclass classification tasks
Expert Syst. Appl.
(2014) - et al.
Logistic regression product-unit neural networks for mapping Ridolfia segetum infestations in sunflower crop using multitemporal remote sensed data
Comput. Electron. Agric.
(2008) - et al.
The use of data mining to assist crop protection decisions on kiwifruit in New Zealand
Comput. Electron. Agric.
(2014)
Wrappers for feature Subset Selection
Artif. Intell.
An intelligent multimedia interface for fuzzy-logic based inference in crops
Expert Syst. Appl.
Optimizing Feature Selection using Particle Swarm Optimization and Utilizing Ventral Sides of Leaves for Plant Leaf Classification Twelfth International Multi-Conference on Information Processing
Random forests ensemble classifier trained with data resampling strategy to improve cardiac arrhythmia diagnosis
Comput. Biol. Med.
Rice diseases classification using feature selection and rule generation techniques
Comput. Electron. Agric.
Tomato crop disease classification using pre-trained deep learning algorithm
Procedia Comput. Sci.
Automatic fruit and vegetable classification from images
Comput. Electron. Agric.
A review of advanced techniques for detecting plant diseases
Comput. Electron. Agric.
Comparative assessment of feature selection and classification techniques for visual inspection of pot plant seedlings
Comput. Electron. Agric.
A systematic analysis of performance measures for classification tasks
Inf. Process. Manage.
An experimental evaluation of ensemble methods for EEG signals classification
Pattern Recogn. Lett.
Automatic identification of music performers with learning ensembles
Artif. Intell.
Computer vision system for on-line sorting of pot plants using an artificial neural network classifier
Comput. Electron. Agric.
Stacked generalization
Neural Networks
Introduction to machine learning
Assessing the accuracy of prediction algorithms for classification and overview
Bioinformatics
An empirical comparison of voting classification algorithms: bagging, boosting, and variants
Mach. Learn.
Bagging predictors
Mach. Learn.
Machine learning classification techniques: a comparative study
Int. J. Adv. Comput. Theory Eng.
Performance evaluation of feature selection methods for Mobile devices
Int. J. Eng. Res. Appl.
Cited by (32)
Application of AI techniques and robotics in agriculture: A review
2023, Artificial Intelligence in the Life SciencesSmart farming using artificial intelligence: A review
2023, Engineering Applications of Artificial IntelligenceComputer vision based method for severity estimation of tea leaf blight in natural scene images
2023, European Journal of AgronomyCollaboration of features optimization techniques for the effective diagnosis of glaucoma in retinal fundus images
2022, Advances in Engineering SoftwareCitation Excerpt :The study [45] was based on real-world Coronary artery disease (CAD) data and attempts to propose a hybrid binary-real PSO that combines categorical and numerical particle encodings and a novel approach for estimating particle velocity. The paper [46] presents a new ensemble approach – Ensemble Particle Swarm Optimization (EnsPSO). The EnsPSO approach is a combination of (i) Vote, (ii) Correlation Based Feature(s) Selection (CFS) method, (iii) PSO algorithm, and (iv) random sampling method.
Citrus greening disease recognition algorithm based on classification network using TRL-GAN
2022, Computers and Electronics in AgricultureCitation Excerpt :Nowadays, deep convolutional models can be used to learn features instead of manual feature extraction. With deep learning technology and increasingly sufficient plant disease image data, the traditional machine learning shows that the problems of extracting plant features and poor generalization ability are effectively solved, and many plant diseases can be recognized simultaneously by using deep convolutional networks instead of traditional manual features extraction (Chaudhary et al., 2020; Ma et al., 2018; Li et al., 2020; Zhou et al., 2021; Waheed et al., 2020; Zeng and Li, 2020; Zhang et al., 2019; Zhang et al., 2021; Gao et al., 2021; Wang et al., 2021; Abade et al., 2021; Jiang et al., 2021; Zhong and Zhao, 2020; Zhang et al., 2019; Gui et al., 2021). Likewise, deep learning has been made a certain achievement in the application of citrus greening disease at this point.