Random cross-validation (CV) splits data into training and testing sets and is widely used to evaluate model performance. However, when the dataset includes multiple records from the same animal, random partitioning will likely create dependencies between the training and testing sets (animal effect) and inflate prediction accuracy. Using a public beef cattle sensor dataset, we examined how different CV designs affect the performance of machine learning and deep learning models for cattle behavior prediction. Our results show that block CV, which splits data by animal, provides a more realistic and reliable evaluation. 🧠 ➡️ 💡  Read the paper here: https://lnkd.in/d5WcNPBY