Integrating phenotypic traits, multivariate statistics, and machine-learning models to predict yield in Himalayan kiwifruit (Actinidia deliciosa) orchards.

BACKGROUND: Understanding the morphological and physiological determinants of kiwifruit (Actinidia deliciosa A. Chev.) yield is essential to improve orchard productivity in Himalayan hill agro-ecosystems. However, most previous studies have focused on individual agronomic factors and have rarely integrated multivariate phenotypic traits with advanced predictive modeling approaches. The aim of this study was to evaluate whether combining multivariate statistical techniques with machine-learning algorithms could improve identification of key traits influencing yield variability in Himalayan kiwifruit orchards. RESULTS: = 0.84; RMSE = 6.20 kg), outperforming gradient boosting and support vector regression. Feature importance analysis consistently identified fruits per shoot, fruit width, vine girth, flowering duration, and leaf-to-fruit ratio as the most influential predictors. CONCLUSION: The integration of multivariate statistical analysis with machine learning provides a robust framework for identifying key trait-based indicators of yield variability. These findings can support clonal selection, optimized pruning strategies, and site-specific orchard management to increase productivity in Himalayan kiwifruit systems.