Scholarlink

<p>Traditional water quality assessment methods often depend on manual sampling and laboratory analysis, which are time-consuming, expensive and impractical for continuous monitoring. However, these struggle with high computational cost and high risk of overfitting. Additionally, existing models also lack interpretability, making it difficult to interpret predictions and enhance trust and transparency. To overcome these challenges, this study proposed a novel Quantized Generative adversarial networks-Convolutional Neural Networks with XAI techniques (QGCNN-X) approach which combines Generative Adversarial Networks (GAN) with CNN as hybrid model for improved feature learning and classification, Quantization for optimization and Explainable AI (XAI) techniques for enhanced interpretability. The observed results indicate that the proposed approach exhibited superior performance in comparison with other conventional Machine Learning (ML) models, generative architectures and optimization techniques. The primary benefits of this proposed solution include environmental agencies, water treatment plants and industries depending on consistent water quality. Overall, this study is important for researchers to understand the importance of enhanced interpretability in environmental sciences.</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

Quantized GAN-based CNN hybrid model with enhanced explainability for water quality assessment

  • 期刊论文

  • 发表日期：

    2025年11月21日

查看全文
  • Gopal Rathinam

摘要

Traditional water quality assessment methods often depend on manual sampling and laboratory analysis, which are time-consuming, expensive and impractical for continuous monitoring. However, these struggle with high computational cost and high risk of overfitting. Additionally, existing models also lack interpretability, making it difficult to interpret predictions and enhance trust and transparency. To overcome these challenges, this study proposed a novel Quantized Generative adversarial networks-Convolutional Neural Networks with XAI techniques (QGCNN-X) approach which combines Generative Adversarial Networks (GAN) with CNN as hybrid model for improved feature learning and classification, Quantization for optimization and Explainable AI (XAI) techniques for enhanced interpretability. The observed results indicate that the proposed approach exhibited superior performance in comparison with other conventional Machine Learning (ML) models, generative architectures and optimization techniques. The primary benefits of this proposed solution include environmental agencies, water treatment plants and industries depending on consistent water quality. Overall, this study is important for researchers to understand the importance of enhanced interpretability in environmental sciences.