Multilayer perceptron and support vector regression models for feline parturition date prediction

Area of Science:

• Veterinary Medicine
• Machine Learning
• Reproductive Biology

Background:

• Accurate prediction of parturition date in feline obstetric care remains a challenge.
• Traditional simple linear regression (SLR) models using fetal biparietal diameter (BPD) have shown limited accuracy.
• Advanced regression models offer potential for improved predictive performance.

Purpose of the Study:

• To introduce and evaluate Multilayer Perceptron (MLP) and Support Vector Regression (SVR) models for feline parturition date prediction.
• To compare the performance of MLP and SVR against traditional methods.
• To identify key input features for enhanced prediction accuracy.

Main Methods:

• Implementation of MLP and SVR models for parturition date prediction.
• Utilized input features: fetal biparietal diameter (BPD), litter size, and maternal weight.
• Comparative analysis of model performance based on coefficient score, mean absolute error, and mean squared error.

Main Results:

• The MLP model demonstrated superior performance with a high coefficient score (0.972 ± 0.006).
• MLP achieved the lowest mean absolute error (1.110 ± 0.060) and mean squared error (1.540 ± 0.141).
• BPD, litter size, and maternal weight were identified as essential features for MLP and SVR models.

Conclusions:

• MLP and SVR represent promising tools for accurate feline parturition date prediction.
• The developed analytical platform and optimized models are suitable for further verification in clinical practice.
• This study establishes a novel approach for improving feline obstetric care through advanced predictive modeling.