Scholarlink

<p>This work highlights the significance of minimizing human intervention for precise precipitation forecasting in Joshimath, Uttarakhand, India, a region prone to heavy rainfall, landslides, and seismic activity. The study uses a correlation matrix to identify relationships between climatic variables and precipitation intensity and develops forecasting models with Automated Machine Learning (AutoML), time series transformers (Informer, Autoformer), and nine traditional ML models (Random Forest (RF), XGBoost, Decision Tree, Bagging, AdaBoost, Ridge Regression, K-Nearest Neighbors, Multiple Linear Regression, and Lasso Regression. Leveraging 14 years of data from Visual Crossing, all models’ performance is evaluated using MAE, RMSE, and <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^2\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation>. The findings indicate that the AutoML (H2O) model does better than others (RMSE: 5.8600, MAE: 1.9787, <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(R^2\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation>: 58.01 %), eliminating the need for manual model selection and hyperparameter tuning, making it a more efficient alternative to traditional machine learning methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improving Weather Prediction in the Himalayas: Automated Machine Learning Approach for Joshimath, Uttarakhand

  • 期刊论文

  • 发表日期：

    2026年7月8日

查看全文
  • Anurag Choubey,
  • Ayush Mittal,
  • Shivendu Mishra,
  • Prince Rajpoot,
  • Bollampalli Areen Reddy,
  • Amit Kumar Pandey,
  • Ritika Yaduvanshi

摘要

This work highlights the significance of minimizing human intervention for precise precipitation forecasting in Joshimath, Uttarakhand, India, a region prone to heavy rainfall, landslides, and seismic activity. The study uses a correlation matrix to identify relationships between climatic variables and precipitation intensity and develops forecasting models with Automated Machine Learning (AutoML), time series transformers (Informer, Autoformer), and nine traditional ML models (Random Forest (RF), XGBoost, Decision Tree, Bagging, AdaBoost, Ridge Regression, K-Nearest Neighbors, Multiple Linear Regression, and Lasso Regression. Leveraging 14 years of data from Visual Crossing, all models’ performance is evaluated using MAE, RMSE, and \(R^2\) R 2 . The findings indicate that the AutoML (H2O) model does better than others (RMSE: 5.8600, MAE: 1.9787, \(R^2\) R 2 : 58.01 %), eliminating the need for manual model selection and hyperparameter tuning, making it a more efficient alternative to traditional machine learning methods.