Scholarlink

<p>In recent years, climate change has intensified, disrupting natural weather patterns and making rainfall prediction increasingly challenging. As rainfall is crucial for determining crop selection and cultivation timing, accurate forecasting can significantly aid farmers in decision-making. This study aims to develop a robust rainfall prediction model using advanced data mining techniques. The proposed approach leverages a stacked ensemble method, integrating multiple weak learners to form a strong predictive model. Specifically, Random Forest (RF), Linear Regression (LR), and K-Nearest Neighbor (KNN) serve as base learners, due to their Complementary characteristics. RF shows good results with non-linear and multidimensional data, K-NN effectively deals with local similarity detection and low complexity datasets and Linear Regression is fast and interpretable for linear trends. Artificial Neural Network (ANN) acts as the meta-model to integrate these diverse outputs, enhancing overall predictive accuracy. It has the capability to model complex interactions and capture non- linear relationships. The model is trained and evaluated using approximately 35 years of historical daily weather data, ensuring reliability through cross-validation. By enhancing rainfall forecasting accuracy, this research seeks to provide farmers with valuable insights for effective agricultural planning amid evolving climatic conditions.</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

Applying stacked ensemble methods to predict future rainfall based on historical weather data

  • 期刊论文

  • 发表日期：

    2025年11月6日

查看全文
  • Deepak Sharma,
  • Preeti Gulia,
  • Nasib Singh Gill,
  • Aliya Aleryani,
  • Prashant Kumar Shukla,
  • Piyush Kumar Shukla

摘要

In recent years, climate change has intensified, disrupting natural weather patterns and making rainfall prediction increasingly challenging. As rainfall is crucial for determining crop selection and cultivation timing, accurate forecasting can significantly aid farmers in decision-making. This study aims to develop a robust rainfall prediction model using advanced data mining techniques. The proposed approach leverages a stacked ensemble method, integrating multiple weak learners to form a strong predictive model. Specifically, Random Forest (RF), Linear Regression (LR), and K-Nearest Neighbor (KNN) serve as base learners, due to their Complementary characteristics. RF shows good results with non-linear and multidimensional data, K-NN effectively deals with local similarity detection and low complexity datasets and Linear Regression is fast and interpretable for linear trends. Artificial Neural Network (ANN) acts as the meta-model to integrate these diverse outputs, enhancing overall predictive accuracy. It has the capability to model complex interactions and capture non- linear relationships. The model is trained and evaluated using approximately 35 years of historical daily weather data, ensuring reliability through cross-validation. By enhancing rainfall forecasting accuracy, this research seeks to provide farmers with valuable insights for effective agricultural planning amid evolving climatic conditions.