Predictive modeling of honey yield in rural apiaries: insight from Chachapoyas, Amazonas, Peru

Abstract

Honey production is influenced by multiple factors, including climatic conditions, hive management practices, and harvest scheduling. This study evaluated the predictive capacity of statistical modeling techniques using data mining algorithms (MARS, CHAID, CART, and Exhaustive) and artificial neural network algorithms (Multilayer Perceptron, MLP) to estimate honey yields in apiaries located in northeastern Peru. A structured survey was conducted with sixty-nine beekeepers across nineteen districts in the Chachapoyas province. Variables included beekeeper experience, instruction, hive count, visit frequency, harvest frequency, additional income-generating activities, and geographic location. Descriptive statistics, non-parametric tests, Spearman correlations, and exploratory factor analysis were applied to identify latent structures. A linear mixed-effects model was used to assess the combined influence of predictors on honey production, with district included as a random effect. Results indicated that hive number, beekeeping experience, harvest frequency, and exclusive engagement in apiculture were statistically associated with increased honey yields. The model explained a substantial proportion of variance, supporting the integration of technical and socio-demographic variables in production forecasting. These findings demonstrate the utility of predictive modeling for informing hive management strategies and improving the operational efficiency of small-scale beekeeping systems in Andean regions.