Design of an Improved Model for AQI Prediction Using Autoencoder-Based Feature Fusion, Hybrid LSTM-CNN, and Reinforcement Learning

SSRN Electronic Journal, Год журнала: 2025, Номер unknown

Опубликована: Янв. 1, 2025

The more severe nature of the urban air pollution demands accurate and proactive methods quality prediction management. state-of-the-art approaches for AQI are typically based on a limited feature set, such as particulate matter basic meteorological variables, tend to ignore complex interactions among various pollutants, biological parameters, factors. Moreover, most developed models designed reactive measures only, with minimal emphasis preemptive prevent degradation in AQI. These limitations motivate research comprehensive, high-dimensional, real-time optimization. Accordingly, this work presents novel multi-model framework involving fusion, hybrid time-series prediction, reinforcement learning control. First, fusion by an Autoencoder combined PCA extracts key latent features from high-dimensional dataset pollutants including PM2.5, NO2, O3, parameters temperature humidity, factors bacterial counts. This reduces dimensionality while retaining 90-95% levels variance within data. A LSTM-CNN model is used next forecasting AQI-related considering their temporal spatial dependencies. achieves accuracy 85-90% up 7 days reducing RMSE 15-20% compared traditional methods. It finally adopts DQN algorithm that can dynamically recommend optimal control strategy, traffic regulation or industrial shutdown, at predictions. optimizes policy 5,000 training episodes reduction 10-15%. proposed offers holistic approach combining temporal-spatial modeling, will yield significant improvement actionable recommendations healthier environments, both technical advancement perspective terms public health.

Язык: Английский

---

Revolutionizing air quality forecasting: Fusion of state-of-the-art deep learning models for precise classification

Urban Climate, Год журнала: 2025, Номер 59, С. 102308 - 102308

Опубликована: Янв. 28, 2025

Язык: Английский

---

Prediction of Air Quality Index (AQI) using machine learning-based tree models—a case study of Gurugram City, India

Theoretical and Applied Climatology, Год журнала: 2025, Номер 156(6)

Опубликована: Май 9, 2025

Язык: Английский

---

Interpretable predictive modeling of non-stationary long time series

Computers & Industrial Engineering, Год журнала: 2024, Номер 194, С. 110412 - 110412

Опубликована: Июль 22, 2024

Язык: Английский

---

Causality-Driven Feature Selection for Calibrating Low-Cost Airborne Particulate Sensors Using Machine Learning

Sensors, Год журнала: 2024, Номер 24(22), С. 7304 - 7304

Опубликована: Ноя. 15, 2024

With escalating global environmental challenges and worsening air quality, there is an urgent need for enhanced monitoring capabilities. Low-cost sensor networks are emerging as a vital solution, enabling widespread affordable deployment at fine spatial resolutions. In this context, machine learning the calibration of low-cost sensors particularly valuable. However, traditional models often lack interpretability generalizability when applied to complex, dynamic data. To address this, we propose causal feature selection approach based on convergent cross mapping within pipeline build more robustly calibrated networks. This in optical particle counter OPC-N3, effectively reproducing measurements PM1 PM2.5 recorded by research-grade spectrometers. We evaluated predictive performance these causally optimized models, observing improvements both while reducing number input features, thus adhering Occam's razor principle. For model, proposed reduced mean squared error test set 43.2% compared model with all SHAP value-based only achieved reduction 29.6%. Similarly, led 33.2% error, outperforming 30.2% selection. By integrating advanced techniques, advances urban quality monitoring, fostering deeper scientific understanding microenvironments. Beyond current cases, method holds potential broader applications other applications, contributing development interpretable robust models.

Язык: Английский

---

Developing a Machine Learning Model Using Satellite Data to Predict the Air Quality Index (AQI) Over Korba Coalfield, Chhattisgarh (India).

Atmospheric Pollution Research, Год журнала: 2024, Номер unknown, С. 102398 - 102398

Опубликована: Дек. 1, 2024

Язык: Английский

---