Generalization Method and System for Recommending Native Bee Species for Meliponiculture Practice

Context: Native stingless bees are essential pollinators for native flora and crops, but their populations decline due to human activities like deforestation and urbanization. Meliponiculture, the breeding of stingless bees for honey production, is a potential solution to the declining populations of these native bees. Problem: While Meliponiculture can contribute to preserving healthy bee colonies, there is currently no straightforward method for potential beekeepers to determine the most suitable native bee species for their specific region. Proposed Solution: This paper proposes and implements a Geographic Information System capable of inferring the suitability of bee species in a given geographical area. IS Theory: The solution aligns with the principles of the General Systems Theory. It adopts a comprehensive approach, considering geographical data, bee occurrences, and satellite imagery as interconnected elements within a unified system. Method: The system employs machine learning and image processing techniques to recommend suitable native bee species for meliponiculture in user-specified locations. It integrates bee occurrence data with satellite images, also considering factors such as land coverage and topography. Regression tree models estimate species viability, and different models were trained with varying features and class balance. Results: The generated maps are used as a basis for a qualitative analysis of the results. Tests with combinations of training parameters demonstrate the system’s capability to effectively generalize the distribution patterns of the species. Contributions and Impact: The proposed system requires only the user’s geolocation for operation and stands as an important tool for potential beekeepers. Meliponiculture expansion directly benefits environmental conservation and enhances pollination for several agricultural crops.