Improving Pest Control with Explainable Artificial Intelligence in Agriculture

Strategic Insights -

Researchers at the Swedish University of Agricultural Sciences have developed an innovative approach to mitigating damage caused by insect pests, which is a significant challenge for farmers worldwide. Traditional methods of using insecticides have led to high synthetic chemical usage, with a large portion of the applied insecticide not reaching its intended target, instead affecting non-target organisms and polluting the environment. The study proposes a method based on explainable artificial intelligence (AI) feature selection and machine learning to detect pests and beneficial insects in field crops.

Key Takeaways:

  • The research aims to reduce damage caused by insect pests in agriculture by developing an AI-powered method for selective application of insecticides, minimizing non-target effects and environmental pollution.
  • The proposed method uses explainable AI feature selection and machine learning to detect pests and beneficial insects in field crops, improving accuracy and reducing computational complexity.
  • The study created an insect-plant dataset reflecting real field conditions, comprising two pest insects (Colorado potato beetle and green peach aphid) and one beneficial insect (seven-spot ladybird).
  • The dataset includes images of insects on different crops (potato, faba bean, and sugar beet) in laboratory and outdoor settings, increasing dataset diversity and broadening the potential application.
  • The researchers compared the proposed method to conventional feature selection techniques, showing improved accuracy (up to 92.62%) and reduced model parameters, memory usage, and training times.
  • The method demonstrates the effectiveness of a simple machine learning algorithm combined with ideal feature selection, achieving robust performance comparable to other methods.
  • The research offers a reliable approach to automatic detection and discrimination of pest and beneficial insects, facilitating the development of alternative pest control approaches that are less harmful to the environment.

Statistics:

  • Improved accuracy: up to 92.62% (Random Forest), 90.16% (Support Vector Machine), 83.61% (KN Nearest Neighbors), and 81.97% (Na & iuml;ve Bayes)
  • Reduced model parameters: 7.22 x 107 (Random Forest), 6.23 x 103 (Support Vector Machine), 3.64 x 104 (KN Nearest Neighbors), and 1.88 x 102 (Na & iuml;ve Bayes)
  • Reduced memory usage: 7.22 x 107 (Random Forest), 6.23 x 103 (Support Vector Machine), 3.64 x 104 (KN Nearest Neighbors), and 1.88 x 102 (Na & iuml;ve Bayes)
  • Reduced training times: approximately half of conventional feature selection techniques

Sources:

  • NewsRx LLC. Researchers at Swedish University of Agricultural Sciences Have Reported New Data on Artificial Intelligence (Improving the Performance of Machine Learning Algorithms for Detection of Individual Pests and Beneficial Insects Using Feature ...). Journal of Engineering. September 1, 2025; p 3086.
  • Improving the Performance of Machine Learning Algorithms for Detection of Individual Pests and Beneficial Insects Using Feature Selection Techniques. Artificial Intelligence In Agriculture, 2025;15(3):377-394.