We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Modeling and forecasting rainfall patterns in India: a time series analysis with XGBoost algorithm.
- Authors
Mishra, Pradeep; Al Khatib, Abdullah Mohammad Ghazi; Yadav, Shikha; Ray, Soumik; Lama, Achal; Kumari, Binita; Sharma, Divya; Yadav, Ramesh
- Abstract
This study utilizes time series analysis and machine learning techniques to model and forecast rainfall patterns across different seasons in India. The statistical models, i.e., autoregressive integrated moving average (ARIMA) and state space model and machine learning models, i.e., Support Vector Machine, Artificial Neural Network and Random Forest Model were developed and their performance was compared against XGBoost, an advanced machine learning algorithm, using training and testing datasets. The results demonstrate the superior accuracy of XGBoost compared to the statistical models in capturing complex nonlinear rainfall patterns. While ARIMA models tend to overfit the training data, state space models prove more robust to outliers in the testing set. Diagnostic checks show the models adequately capture the time series properties. The analysis indicates essential unchanging rainfall patterns in India for 2023–2027, with implications for water resource management and climate-sensitive sectors like agriculture and power generation. Overall, the study highlights the efficacy of modern machine learning approaches like XGBoost for forecasting complex meteorological time series. The framework presented enables rigorous validation and selection of optimal techniques. Further applications of such sophisticated data analysis can significantly enhance planning and research on the Indian monsoons amidst climate change challenges.
- Subjects
INDIA; TIME series analysis; ARTIFICIAL neural networks; RAINFALL; WATER management; MACHINE learning; BOX-Jenkins forecasting
- Publication
Environmental Earth Sciences, 2024, Vol 83, Issue 6, p1
- ISSN
1866-6280
- Publication type
Article
- DOI
10.1007/s12665-024-11481-w