Prediction of Humidity Based on CEEMDAN-LSTM Approach

• Published in: 2025 International Conference on Frontier Technologies and Solutions (ICFTS) pp. 1 - 7
• Main Authors: S, Daphin Lilda, R, Jayaparvathy
• Format: Conference Proceeding
• Language: English
• Published: IEEE 27.03.2025
• Subjects: Accuracy; Adaptation models; Biodiversity; Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN); Data models; Humidity; Humidity Prediction; Long short term memory; Long Short-Term Memory (LSTM); Noise; Predictive models; Prosopis juliflora; Time series analysis; Vegetation
• DOI: 10.1109/ICFTS62006.2025.11032059

The introduction of Prosopis juliflora in India was a remedy to a rising need for fuel. But over the years, these species have invaded over a million hectares creating an adverse impact on the biodiversity affecting both agriculture and the native plant life. The juliflora are found to increase the humidity levels to a great extent if they are densely spread. Thus, accurate prediction of the humidity in regions where the juliflora are present is required for improved resource management, decision-making, and adaptability to environmental changes. This study suggests an innovative approach based on the CEEMDAN decomposition technique and the LSTM model for humidity prediction in the presence of these noninvasive species. The proposed work for the prediction is validated using the statistical metrics namely, mean absolute error (MAR), mean square error (MSE), R-squared value, and root mean square error (RMSE). From the various metrics, it is concluded that the proposed CEEMDAN combined with the LSTM network can accurately predict humidity with an R-squared value of 0.8686, which was found to be higher compared to the prediction made through the LSTM model. Maximizing the cultivation of the native species by proper pattern and scheduling based on the future humidity values forecast by the suggested CEEMDAN-LSTM method may contribute to the expansion of the green cover in the vicinity.