Time-Series Neural Network Predictions Using LSTM for Clustering and Forecasting GPS Data of Bird Immigration

Abstract

The study addresses the challenge of clustering and forecasting GPS data of bird immigration using time-series neural network predictions with Long Short-Term Memory (LSTM). This approach leverages the unique ability of LSTM to capture temporal patterns in the data, making it well-suited for analyzing the spatio-temporal nature of bird migration. Additionally, the integration of the ST-DBSCAN algorithm for clustering further enhances the model's predictive capabilities by identifying distinct groups within the data. The code's key steps involve data pre-processing, clustering, model creation, and training, culminating in the evaluation of model performance. The results demonstrate the decreasing loss function and metric values over time for both training and validation sets, indicating the model's ability to learn from the training data. Overall, this approach offers a promising method for accurate clustering and forecasting of GPS data of bird immigration, with implications for understanding migration patterns and ecological research.