Animal Movement Prediction Based on Predictive Recurrent Neural Network.

Area of Science:

• Ecology
• Computational Biology
• Animal Behavior

Background:

• Animal movement is intrinsically linked to ecological factors like weather and seasons, influencing behaviors such as foraging, nesting, and migration.
• Understanding animal movement patterns provides critical ecological insights and reveals species-specific traits.
• Predicting these movements is essential for conservation and ecological research.

Purpose of the Study:

• To develop and validate a predictive scheme for animal movement patterns.
• To leverage recurrent neural network architecture for spatiotemporal movement prediction.
• To analyze the relationship between animal traits and their movement behaviors.

Main Methods:

• Collection and investigation of animal geolocational data.
• Pattern refinement using random forest interpolation.
• Generation of movement patterns via kernel density estimation.
• Development of a predictive recurrent neural network model incorporating spatiotemporal dynamics.

Main Results:

• The recurrent neural network model demonstrated effectiveness in predicting animal movements.
• Experimental validation using long-billed curlew movement data across different seasons confirmed predictive accuracy.
• The model successfully captured complex spatiotemporal variations in animal trajectories.

Conclusions:

• Recurrent neural networks offer a powerful tool for predicting animal movement patterns.
• The proposed scheme provides a robust method for acquiring ecological insights through movement analysis.
• Accurate animal movement prediction can significantly aid ecological studies and conservation efforts.