Modelling of pome fruit pollen performance using machine learning

Area of Science:

• Agricultural Science
• Bioinformatics
• Machine Learning in Agriculture

Background:

• Extreme temperature fluctuations negatively impact fruit crop flowering and yield.
• Pollen performance analysis is vital for understanding plant reproductive success.
• Advancements in technology have driven new methods for pollen germination analysis.

Purpose of the Study:

• To develop a predictive model for pollen germination rates in pome fruits using machine learning.
• To assess the influence of environmental factors on pollen viability and germination.
• To leverage artificial intelligence for improved agricultural research outcomes.

Main Methods:

• In vitro testing of pollen germination rate and pollen tube length for four pome fruit cultivars.
• Experimentation across three different media, four incubation durations, and seven temperature variations.
• Development and validation of three deep learning models with two hidden layers, evaluating various optimizers.

Main Results:

• A machine learning model, specifically using artificial neural networks with the Adam optimizer, achieved a high R² value of 0.89.
• The developed model demonstrated significant accuracy in predicting pollen germination rates.
• The study successfully identified key factors influencing pollen performance under varied conditions.

Conclusions:

• Machine learning, particularly deep learning, offers a powerful tool for predicting pollen germination rates in agriculture.
• Accurate prediction of pollen viability can aid in optimizing fruit production and mitigating climate change impacts.
• This research underscores the potential of AI to advance agricultural science and improve crop management strategies.