Modelling of pome fruit pollen performance using machine learning

Agriculture, particularly fruit production, is considered a crucial industry with a significant economic impact in many countries. Extreme fluctuations in air temperature can negatively affect the flowering periods of fruit species. Therefore, it is important to conduct studies on pollen performance analysis to determine these effects. Pollen performance analysis has seen significant advancements in agricultural research, with the emergence of new pollen germination methods driven by advances in technology and equipment. In these analyses, in addition to traditional approaches, Artificial Neural Networks and deep learning have gained importance recently. The main objective of this study is to develop a model that predicts the germination rate based on a given set of input variables. Firstly, pollen germination rate and pollen tube length were determined in vitro. Pollen grains from four cultivars of pome fruit were sown in three different media and incubated for four different durations at seven different temperatures for an in vitro test. Three deep-learning models with two hidden layers were developed and different optimizers were considered for model development. The best model was selected through a validation test.This study aimed to develop a machine learning model for predicting pollen germination rates in pome fruits. Pollen grains from four cultivars were subjected to in vitro germination tests under varying temperatures, media, and durations. Using artificial neural networks, the model achieved an R² value of 0.89 with the Adam optimizer, demonstrating high accuracy in predicting germination rates. These findings highlight the potential of machine learning in advancing agricultural research.