Novel hybrid data-driven modeling based on feature space reconstruction and multihead self-attention gated recurrent unit: applied to PM2.5 concentrations prediction DOI Creative Commons
Xin Yue, Yulong Bai,

Qinghe Yu

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 16, 2025

In response to the problem of neglecting periodic and global characteristics sequence data when predicting PM2.5 concentrations via machine learning models, a prediction model based on feature space reconstruction multihead self-attention gated recurrent unit (FSR-MSAGRU) is proposed in this study. First, raw are subjected frequency spectrum analysis determine period value data. Subsequently, seasonal trend decomposition procedure loess (STL) employed capture periodicity information Then, reconstructed using data, decomposed components, residual components. Finally, input into (MSAGRU) with ability predict concentrations. Favorable results were attained by FSR-MSAGRU across 6 distinct experimental datasets, PCC exceeding 0.98 decrease accuracy metric SMAPE at least 68% compared that GRU model. Comparative 13 reference models demonstrate exhibits better performances stronger generalization abilities.

Язык: Английский

Novel hybrid data-driven modeling based on feature space reconstruction and multihead self-attention gated recurrent unit: applied to PM2.5 concentrations prediction DOI Creative Commons
Xin Yue, Yulong Bai,

Qinghe Yu

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 16, 2025

In response to the problem of neglecting periodic and global characteristics sequence data when predicting PM2.5 concentrations via machine learning models, a prediction model based on feature space reconstruction multihead self-attention gated recurrent unit (FSR-MSAGRU) is proposed in this study. First, raw are subjected frequency spectrum analysis determine period value data. Subsequently, seasonal trend decomposition procedure loess (STL) employed capture periodicity information Then, reconstructed using data, decomposed components, residual components. Finally, input into (MSAGRU) with ability predict concentrations. Favorable results were attained by FSR-MSAGRU across 6 distinct experimental datasets, PCC exceeding 0.98 decrease accuracy metric SMAPE at least 68% compared that GRU model. Comparative 13 reference models demonstrate exhibits better performances stronger generalization abilities.

Язык: Английский

Процитировано

0