Scholarlink

<p>This paper designs an integrated framework using machine learning and ensemble models for single and multi-label classification of crops. For both types of classification, we first applied Gaussian Naive Bayes, logistic regression, multilayer perceptron, K-nearest neighbors, decision tree, random forest, and hoeffding tree individually. Next, different ensemble models, including bagging, voting, and stacking, were implemented and evaluated. Bagging classifiers for Decision Trees were tuned for the number of estimators and maximum sample size using a grid search approach. On the other hand, voting classifiers (hard and soft) that combine two weak models with one strong model, and stacking classifiers that use weak models as base models and a strong model as the meta-model, were trained using default hyperparameters. It has been found from the results that for single-label classification, Gaussian Naive Bayes achieved the best accuracy of 99.44% with a kappa of 0.994, mean absolute error of 0.0661, root mean squared error of 0.890, and a time complexity of 0.01 s. The best bagging classifier gave 99.17% accuracy. Voting classifiers reached an accuracy of 99.44% (hard) and 98.89% (soft), while the best stacking model gave 98.07%. In multi-label classification, K-nearest neighbors gave 97.82% accuracy. Bagging and voting achieved accuracies of 98.67% and 98.87%, respectively, and the highest-performing stacking classifiers demonstrated an accuracy of 94.82%. In conclusion, this research examined a variety of accurate and efficient models that help improve productivity and economic growth in agriculture.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An integrated ensemble approach using diverse machine learning models for single-label and multi-label crop classification

  • 期刊论文

  • 发表日期：

    2026年3月9日

查看全文
  • Dilpreet Kaur,
  • Rohit Kumar Singla,
  • Kavita Goyal

摘要

This paper designs an integrated framework using machine learning and ensemble models for single and multi-label classification of crops. For both types of classification, we first applied Gaussian Naive Bayes, logistic regression, multilayer perceptron, K-nearest neighbors, decision tree, random forest, and hoeffding tree individually. Next, different ensemble models, including bagging, voting, and stacking, were implemented and evaluated. Bagging classifiers for Decision Trees were tuned for the number of estimators and maximum sample size using a grid search approach. On the other hand, voting classifiers (hard and soft) that combine two weak models with one strong model, and stacking classifiers that use weak models as base models and a strong model as the meta-model, were trained using default hyperparameters. It has been found from the results that for single-label classification, Gaussian Naive Bayes achieved the best accuracy of 99.44% with a kappa of 0.994, mean absolute error of 0.0661, root mean squared error of 0.890, and a time complexity of 0.01 s. The best bagging classifier gave 99.17% accuracy. Voting classifiers reached an accuracy of 99.44% (hard) and 98.89% (soft), while the best stacking model gave 98.07%. In multi-label classification, K-nearest neighbors gave 97.82% accuracy. Bagging and voting achieved accuracies of 98.67% and 98.87%, respectively, and the highest-performing stacking classifiers demonstrated an accuracy of 94.82%. In conclusion, this research examined a variety of accurate and efficient models that help improve productivity and economic growth in agriculture.