Machine learning based prediction of hexapod invertebrates and its impact on biodiversity

Abstract

Insects represent one of the most diversified groups and are also critical to maintaining ecological balance; however, their monitoring is limited by traditional field sampling and manual identification techniques. This research introduces a new framework that combines automated insect detection with biodiversity quantification and hotspot visualization, converting the recognition results into ecological conservation decision-support tools. We have used a dataset of 6,000 annotated images of ten equivalent classes of hexapods to train and evaluate the state-of-the-art systems, YOLOv8 and YOLOv10, in a hybrid setup. The YOLOv8 model showed better results in all the tested detection metrics. Biodiversity was also determined by ecological metrics besides detection accuracy, which also showed that the site had a species richness of 10, a Shannon diversity index of 2.296 and an evenness value of 0.997. To find diversity gradients across locations, numerical indices were replaced with a hotspot analysis. The maps indicate sites that need to be preserved right away, not places that can be restored environmentally. The study shows that combining advanced deep learning methods with biodiversity measurements can lead to data-rich, scalable, and ecologically sound insights for monitoring and making policy decisions.