A Perspective of Mango Pest Classification Using Wrapper-Based Feature Selection Algorithm

The presence of pests on mango fruits reduces productivity and increases the need for pesticide usage. Therefore, early pest identification can significantly impact productivity. However, identifying various types of pests from images is a challenging and time-consuming process.This article introduces a feature selection technique using a wrapper approach to improve classification accuracy. The method involves adapting the basic random forest classifier by choosing a subset of high-accuracy features to input into the classifier. Our proposed modified random forest (MRF) classifier achieves a classification accuracy of 99.8%, representing a substantial improvement of 7.19%, 13.53%, and 7.65% compared to the random forest, k-nearest neighbor, and Ada Boost classifiers, respectively.