Environment and human health are seriously threatened by air pollution. The effects of air pollution are more severe in metropolitan areas due to the presence of harmful pollutants. The goal of this work is to forecast the Air Quality Index (AQI), of 15 metropolitan cities in India and analyze various air pollutants that are mostly responsible for higher levels of air pollution in a particular city. Firstly, air quality data from 15 metropolitan cities were gathered and preprocessed appropriately. The prediction models were then trained using the preprocessed dataset. Modified Particle Swarm Optimization (MPSO)-based two hybrid deep learning models: Long-Short Term Memory (LSTM) along with Bi-directional Recurrent Neural Network (BiRNN) and LSTM along with Gated Recurrent Unit (GRU) are proposed and the experimental analysis demonstrated that the proposed MPSO-LSTM-BiRNN and MPSO-LSTM-GRU models outperformed the other models' performance in terms of Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) values. MPSO-LSTM-BiRNN model provides MSE, RMSE, MAE, and MAPE of 0.000184, 0.0135, 0.0088, and 27.69 % respectively whereas, the MPSO-LSTM-GRU model gives MSE, RMSE, MAE, and MAPE of 0.000188, 0.0137, 0.0091 and 26.16 % respectively.

中文翻译：

---

一种改进的基于 PSO 的混合深度学习方法预测城市都市 AQI


空气污染严重威胁着环境和人类健康。由于有害污染物的存在，空气污染的影响在大都市地区更为严重。这项工作的目标是预测印度 15 个大城市的空气质量指数 （AQI），并分析导致特定城市空气污染水平较高的各种空气污染物。首先，收集了来自 15 个大城市的空气质量数据并进行了适当的预处理。然后使用预处理后的数据集对预测模型进行训练。提出了基于改进的粒子群优化 （MPSO） 的两种混合深度学习模型：长短期记忆 （LSTM） 以及双向递归神经网络 （BiRNN） 和 LSTM 以及门控循环单元 （GRU），实验分析表明，所提出的 MPSO-LSTM-BiRNN 和 MPSO-LSTM-GRU 模型在均方误差 （MSE） 方面优于其他模型的性能， 均方根误差 （RMSE）、平均绝对误差 （MAE） 和平均绝对百分比误差 （MAPE） 值。MPSO-LSTM-BiRNN 模型分别提供 0.000184、0.0135、0.0088 和 27.69% 的 MSE、RMSE、MAE 和 MAPE，而 MPSO-LSTM-GRU 模型分别给出 0.000188、0.0137、0.0091 和 26.16% 的 MSE、RMSE、MAE 和 MAPE。