跳转至

使用 QHAdamW 优化空气质量预测中的增强型人工神经网络

文章背景与核心概要

本研究提出了一种结合了拟双曲动量（QHAdam）与解耦权重衰减的 Adam（AdamW）优化器的增强型人工神经网络（ANN）模型，旨在显著提升空气质量指数（AQI）的预测精度。通过利用马尼拉实时空气质量监测站的数据，该前馈神经网络被部署用于分别预测细颗粒物（\(\text{PM}_{2.5}\) 和 \(\text{PM}_{10}\)）。

研究所提出的 QHAdamW 优化器成功克服了传统模型在收敛速度、泛化能力以及预测准确性方面的局限性。超参数调优结果表明，0.01 和 0.001 是实现该优化器最佳泛化性能的最优值。通过七种评估指标的对比分析证实，该模型具有更低的误差范围、接近 1 的回归系数以及更低的训练和验证损失值，为菲律宾环境与自然资源部环境管理局（DENR-EMB）的环境监测与空气质量管理提供了一个强有力的技术工具。

📌 Summary

This study presents an enhanced Artificial Neural Network (ANN) model leveraging the QHAdamW optimizer—a combination of Quasi-Hyperbolic Momentum (QHAdam) and Adam with decoupled weight decay (AdamW)—to improve Air Quality Index (AQI) forecasting. Utilizing real-time data from a tracking station in Manila, a feed-forward neural network was deployed to predict particulate matter (\(\text{PM}_{2.5}\) and \(\text{PM}_{10}\)) separately. The proposed optimizer successfully addresses traditional limitations in convergence, generalization, and model accuracy, providing a robust tool for environmental monitoring and air quality management by the Philippine Department of Environment and Natural Resources-Environmental Monitoring Bureau (DENR-EMB).

📄 Article Overview

  • Title: Enhanced Artificial Neural Networks Using QHAdamW in Air Quality Forecasting
  • Author: Mary Joy Daniel Vinas
  • Submitted Date: August 20, 2026
  • Primary Subject: Artificial Intelligence (cs.AI)
  • Secondary Subjects: Emerging Technologies (cs.ET), Machine Learning (cs.LG), Neural and Evolutionary Computing (cs.NE)
  • Journal Reference: International Journal of Engineering Trends and Technology, 74(7), 352–371
  • DOI: 10.48550/arXiv.2608.21463
  • Related DOI: 10.14445/22315381/IJETT-V74I7P122

🔍 Abstract

本研究采用人工神经网络结合优化后的自适应矩估计（Adam）算法，这是目前菲律宾唯一可用的 AQI 预测模型。改进后的 QHAdamW 融合了拟双曲动量（QHAdam）与解耦权重衰减的 Adam（AdamW），这两者皆为 Adam 优化器的扩展，能为人工神经网络的训练提供独特的优势。

The study employed an Artificial Neural Network in combination with the optimized Adaptive Moment Estimation (Adam) algorithm, currently the only AQI forecasting model available in the Philippines. The modified QHAdamW—Quasi-Hyperbolic Momentum (QHAdam) and Adam with decoupled weight decay (AdamW)—were both extensions of the Adam optimizer, and both offer unique advantages for training ANN.

所提出的 QHAdamW 优化器有效解决了 Adam 在收敛性、泛化能力以及预测性能方面的问题。超参数调优结果显示，0.01 和 0.001 是 QHAdamW 获得最佳泛化性能的最有效参数值。

The proposed QHAdamW optimizer addresses the issues on convergence, generalization, and forecasting performance of Adam. Hyperparameter tuning results revealed that 0.01 and 0.001 were the most effective optimal values for the generalization performance of QHAdamW.

基于七个评估指标的对比分析结果表明： * 误差值范围更低。 * 回归系数值约等于 \(1\)，从而提高了模型准确性。 * 模型收敛至令人满意的性能水平，这体现在更低的训练和验证损失值上。

The comparative analysis results using seven evaluation metrics revealed that: * The error value range is lower. * The regression coefficient has a value approximately equal to \(1\), improving model accuracy. * The model converges to a satisfactory level of performance, demonstrated by lower training and validation loss values.

利用马尼拉实时空气质量追踪站的数据，前馈神经网络被用于分别预测 \(\text{PM}_{2.5}\) 和 \(\text{PM}_{10}\) 的 AQI。该模型可用于预测颗粒物（PM），从而协助菲律宾环境与自然资源部环境管理局（DENR-EMB）实施全面的空气质量管理。

Based on data from a real-time air quality tracking station in Manila, a feed-forward neural network is used to predict the AQI of \(\text{PM}_{2.5}\) and \(\text{PM}_{10}\) separately. This model can be used to forecast Particulate Matter (PM), to help the Department of Environment and Natural Resources-Environmental Monitoring Bureau (DENR-EMB) implement comprehensive air quality management.

🔗 Access & Resources

访问链接与资源：