Scholarlink

<p>This paper introduces an innovative Gradient Light Spectrum Optimization-based Deep Long Short-Term Memory (GLSO-DLSTM) model. The key innovation lies in the integration of Gradient Light Spectrum Optimization (GLSO), a hybrid optimization technique combining Gradient-based Dragonfly Optimization (GDO) and Light Spectrum Optimization (LSO), for effective feature weighting and hyperparameter tuning. Z-score normalization is applied to preprocess the input data, after which the Synthetic Minority Over-sampling Technique for Regression (SMOTE-REG) is used to address data imbalance. Following augmentation, the weighted features are computed using the proposed GLSO algorithm. Water quality prediction is ultimately performed using the DLSTM optimized by GLSO. Further, the proposed technique is analysed for its efficiency based on Mean Absolute Error (MAE), Mean Square Error (MSE), Median Absolute Error (MedAE), and coefficient of determination (R<sup>2</sup>), which are found to attain superior values of 0.136, 0.002, 0.099, and 0.969, respectively.</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

Gradient Light Spectrum Optimization based Deep Long Short-Term Memory for Water Quality Prediction

  • 期刊论文

  • 发表日期：

    2026年6月30日

查看全文
  • Dhanalakshmi Lanka,
  • Jaya Kumari D,
  • Ramakrishnaiah N

摘要

This paper introduces an innovative Gradient Light Spectrum Optimization-based Deep Long Short-Term Memory (GLSO-DLSTM) model. The key innovation lies in the integration of Gradient Light Spectrum Optimization (GLSO), a hybrid optimization technique combining Gradient-based Dragonfly Optimization (GDO) and Light Spectrum Optimization (LSO), for effective feature weighting and hyperparameter tuning. Z-score normalization is applied to preprocess the input data, after which the Synthetic Minority Over-sampling Technique for Regression (SMOTE-REG) is used to address data imbalance. Following augmentation, the weighted features are computed using the proposed GLSO algorithm. Water quality prediction is ultimately performed using the DLSTM optimized by GLSO. Further, the proposed technique is analysed for its efficiency based on Mean Absolute Error (MAE), Mean Square Error (MSE), Median Absolute Error (MedAE), and coefficient of determination (R2), which are found to attain superior values of 0.136, 0.002, 0.099, and 0.969, respectively.