基于SSA-LSTM模型的空气质量预测研究
Research on Air Quality Prediction Based on SSA-LSTM Model
曹还君 1李长云1
作者信息
- 1. 湖南工业大学 计算机学院,湖南 株洲 412007
- 折叠
摘要
为提高PM2.5 浓度的预测精度,提出了一种结合麻雀搜索算法(SSA)和长短期记忆神经网络(LSTM)的组合预测模型.以 2023 年 5 月至 8 月期间长沙市PM2.5 浓度数据为基础,构建了SSA-LSTM模型并与其他模型进行了对比实验.实验结果显示,SSA-LSTM模型的预测结果在拟合优度(R2)上相较于单一LSTM、PSO-LSTM和WOA-LSTM模型分别提升了 45.93％、31.55％、19.12％,同样在均方根误差(RMSE)和平均绝对误差(MAE)的结果上也表现更优,表明该模型在PM2.5 浓度预测方面具有高准确性和有效性,可为制定PM2.5 相关预防措施提供一定的参考价值.
Abstract
To improve the accuracy of PM2.5 concentration prediction,a combined prediction model integrating Sparrow Search Algorithm(SSA)and Long Short-Term Memory(LSTM)neural networks is proposed.The SSA-LSTM model is developed based on PM2.5 concentration data from Changsha city,spanning from May to August in 2023,and is compared with other models.The results show that the SSA-LSTM model significantly outperformed the standalone LSTM,PSO-LSTM,and WOA-LSTM models in terms of fit quality(R2),registering improvements of 45.93％,31.55％,and 19.12％,respectively.Similarly,it also shows superior performance in terms of Root Mean Square Error(RMSE)and Mean Absolute Error(MAE).These findings demonstrate the model has high accuracy and effectiveness in PM2.5 concentration prediction,providing a certain reference value for making the PM2.5-related preventive measures.
关键词
麻雀搜索算法/长短期记忆神经网络/空气质量/PM2.5浓度预测Key words
SSA/LSTM/air quality/PM2.5 concentration prediction引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据