The Greater Bushenyi region of southwest Uganda, rich in both ecological and cultural diversity, faces persistent challenges of poverty and malnutrition. In this region, banana-based homegardens (HGs) dominate the landscape, having largely replaced the native Guineo-Congolian rainforest. These HGs are crucial for household food security. To explore pathways for improving nutrition through agrobiodiversity, a study was conducted across 102 HGs to classify garden types based on plant diversity. Using Ward’s minimum variance hierarchical clustering (based on Mahalanobis and squared Euclidean distances of log-transformed useful plant densities per 1,000 m²), four distinct HG types emerged. Clusters 1 and 2 exhibited the highest Shannon diversity (H’) and evenness (Pielou’s J’), while Cluster 3 had the greatest species richness. Cluster 4 stood out for its abundance of annual and perennial herbs and the overall highest plant densities. Linear discriminant analysis (LDA) achieved 98% classification accuracy, and a random forest model explained 51% of variance in cluster assignments. Thirteen plant species significantly contributed to the cluster model, with eight species—Phaseolus vulgaris, Coffea arabica, Amaranthus dubius, Amaranthus hybridus, Saccharum officinarum, Amaranthus spinosus, Solanum torvum, and Artocarpus heterophyllus—having the lowest mean squared errors (MSE <1.3%). Stepwise selection further identified eleven species strongly associated with HG type (p<0.001). These findings suggest that enhancing the nutritional quality of homegardens should begin with a targeted, diversity-based approach rooted in local agrobiodiversity and garden typologies. Interventions that focus on key species within specific HG types could improve food and nutrition security in the region.