Harnessing ensemble models for improved lactation curve modeling: The R package EMOTIONS.

Area of Science:

• Animal Science
• Dairy Production
• Statistical Modeling

Background:

• Lactation curve modeling is crucial for dairy production, influencing management and breeding decisions.
• Accurately modeling diverse individual lactation patterns poses a significant challenge.
• Existing single models often struggle to capture the full spectrum of lactation variability.

Purpose of the Study:

• To introduce a flexible ensemble modeling (EM) framework, EMOTIONS R package, for improved milk yield prediction in dairy animals.
• To enhance the prediction of daily milk yield and related traits in cows and ewes.
• To provide tools for robust lactation data analysis, including imputation and resilience assessment.

Main Methods:

• Developed the EMOTIONS R package integrating multiple lactation curve models.
• Implemented ensemble creation by weighting individual model predictions based on various criteria.
• Utilized statistical criteria (e.g., BIC) to assess individual model fits and identify areas where ensembles excel.

Main Results:

• Ensemble models demonstrated clear advantages in predicting milk yield, particularly for animals with poor individual model fits.
• EMs effectively captured individual lactation variation and reduced underfitting issues in both cow and ewe datasets.
• The EMOTIONS package offers comprehensive tools for fitting, ensemble generation, visualization, and data imputation.

Conclusions:

• Ensemble modeling provides a robust and adaptable approach for analyzing complex lactation data.
• EMs offer superior predictive accuracy compared to single models, especially for heterogeneous populations.
• This framework supports better on-farm decision-making and the selection of resilient, high-producing dairy animals.