Scholarlink

<p>Rainfall is a natural climatic factor that plays an important role in the economy of a nation by improving agricultural production, power generation, construction, tourism, and forestry. Additionally, in adverse conditions of weather, rainfall is related to floods, avalanches, landslides, and mass movements. Hence, to avoid these worst conditions, prediction of rainfall at an earlier stage seems to be necessary. In this research, rainfall is predicted using a DL technique named ConvLSTM, whose parameters are optimized by the proposed HCDDO algorithm. Here, time-series data is taken as input data, from which the technical indicators are extracted. Moreover, feature fusion is done by MI and Rider Optimization Algorithm (ROA) based Neural Network (RideNN). Then, data augmentation is done by oversampling, followed by rainfall prediction using ConvLSTM, which is trained by HCDDO. The supremacy of HCDDO_ConvLSTM is found based on various performance measures, like MSE, RAE, MAPE, and RMSE, and the experimentation results demonstrate that HCDDO_ConvLSTM attained a low value of MAPE, MSE, RAE, RMSE and R-Squared of 0. 302, 0.255, 0.523, 0.505, and 0.569, correspondingly.</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

Deep learning-driven rainfall prediction leveraging hybrid child drawing development optimization and time series data

  • 期刊论文

  • 发表日期：

    2025年2月22日

查看全文
  • Nihar Ranjan,
  • Mubin Tamboli,
  • Jayashree R. Prasad,
  • Rajesh S. Prasad,
  • Amol V. Dhumane

摘要

Rainfall is a natural climatic factor that plays an important role in the economy of a nation by improving agricultural production, power generation, construction, tourism, and forestry. Additionally, in adverse conditions of weather, rainfall is related to floods, avalanches, landslides, and mass movements. Hence, to avoid these worst conditions, prediction of rainfall at an earlier stage seems to be necessary. In this research, rainfall is predicted using a DL technique named ConvLSTM, whose parameters are optimized by the proposed HCDDO algorithm. Here, time-series data is taken as input data, from which the technical indicators are extracted. Moreover, feature fusion is done by MI and Rider Optimization Algorithm (ROA) based Neural Network (RideNN). Then, data augmentation is done by oversampling, followed by rainfall prediction using ConvLSTM, which is trained by HCDDO. The supremacy of HCDDO_ConvLSTM is found based on various performance measures, like MSE, RAE, MAPE, and RMSE, and the experimentation results demonstrate that HCDDO_ConvLSTM attained a low value of MAPE, MSE, RAE, RMSE and R-Squared of 0. 302, 0.255, 0.523, 0.505, and 0.569, correspondingly.