An Empirical Mode Decomposition Fuzzy Forecast Model for Air Quality

Wenxin Jiang, Guo-Chang Zhu, Yi-Yun Shen, Qian Xie, Min Ji, Yongtao YuEntropy 2022 · DOI

TL;DR

This paper introduces a novel air quality forecasting model that combines Empirical Mode Decomposition (EMD) with Extreme Learning Machines (ELM) and an Adaptive Fuzzy Inference System (ANFIS). The model effectively captures complex temporal patterns in air quality data, leading to significantly improved prediction accuracy (around 30% better) compared to traditional methods for both short-term and long-term forecasts.

What Problem Does This Paper Address?

Accurate air quality prediction is crucial for public health and environmental management. Air pollutant concentrations form complex, non-linear, and non-stationary time series. Traditional neural network models often struggle with these characteristics, frequently getting stuck in local optima during training and exhibiting slow convergence. Existing methods often focus on a limited set of pollutants, neglecting the broader picture of air quality dynamics. This paper aims to address these limitations by developing a more robust and accurate forecasting model that can handle the inherent complexities of air quality data.

Methods

The core of the proposed model, the Empirical-Mode-Decomposition-Based Fuzzy Prediction Model (EMD-FPM), involves a three-stage process:

  1. Empirical Mode Decomposition (EMD): This technique is used to decompose the complex air quality time series into a set of simpler, more regular intrinsic mode functions (IMFs). Each IMF represents a different time scale or frequency component of the original signal, effectively smoothing out noise and revealing underlying trends. This is a key differentiator from standard time-series models that might not explicitly handle these multi-scale variations.
  2. Extreme Learning Machine (ELM): The IMFs generated by EMD are then fed into an ELM. ELM is chosen for its rapid training speed and ability to avoid local optima, which are common issues with traditional back-propagation neural networks. It's particularly effective for learning from the decomposed components.
  3. Adaptive Fuzzy Inference System (ANFIS): Finally, the outputs from the ELM are integrated using an ANFIS. This system fuses the predictions from the different IMFs, leveraging fuzzy logic to capture non-linear relationships and produce a final, refined air quality prediction. This adaptive fusion step is crucial for achieving high accuracy and generalization.

Datasets

The study utilized air quality data from six monitoring stations in Huaian, Jiangsu province, China. The dataset included daily measurements for AQI, PM2.5, PM10, SO2, NO2, CO, and O3 for the year 2019. This dataset was preprocessed and used for both training and evaluating the proposed EMD-FPM model, as well as for comparison with other models. No new benchmark dataset was introduced.

Key Insights and Ideas

  • Decomposition is Key: The paper strongly suggests that breaking down complex air quality signals into their constituent time scales (IMFs) is essential for accurate forecasting. This allows models to better understand and predict the behavior at different temporal resolutions.
  • ELM's Speed Advantage: The choice of ELM for processing the IMFs is a strategic one. Its ability to achieve good performance with a single training pass and avoid local optima makes it a powerful tool for handling the numerous components generated by EMD.
  • Fuzzy Logic for Refinement: ANFIS acts as a sophisticated "corrector" or "integrator," taking the ELM's predictions for each IMF and intelligently combining them to produce a more accurate final output. This highlights the benefit of hybrid approaches that combine different AI paradigms.
  • Pollution Control Dynamics: The analysis of the SO2 decomposition results hints at the challenges of air quality control, suggesting that temporary measures may lead to temporary improvements, but sustained efforts are needed to prevent pollution recurrence.

Key Results

The EMD-FPM model demonstrated superior performance compared to traditional models like BP neural networks, ELM, and NAR (Nonlinear Auto-Regressive) neural networks.

  • Accuracy Improvement: The proposed model achieved approximately 30% higher prediction accuracy across various air quality indicators for both short-term (predicting the 7th day using 6 days of data) and long-term (predicting days 13-15 using 12 days of data) forecasts.
  • Evaluation Metrics: Across all evaluation metrics (MSE, MAE, SD), the EMD-FPM consistently outperformed the other models, indicating lower errors and better dispersion of predictions.
  • Adaptability: The paper highlights that the EMD-FPM shows good adaptability, even in complex and changeable situations, suggesting its robustness.

Limitations and Future Work

The authors acknowledge several limitations:

  • Limited Sample Size and Region: The study was conducted using data from a single city (Huaian) and a limited sample size (one year). This might limit the generalizability of the findings.
  • Ignoring External Factors: The study primarily focused on historical air quality data and did not explicitly incorporate the impact of air pollution from surrounding cities or detailed seasonal weather conditions.

Future research directions suggested include:

  • Expanding Datasets: Incorporating data from multiple regions and a larger sample size to validate the model's performance more broadly.
  • Considering Inter-City Pollution: Investigating how pollution from neighboring cities influences the target region's air quality.
  • Integrating Weather Data: Including detailed seasonal weather patterns to further enhance prediction accuracy.

Summary updated

Abstract

Air quality has a significant influence on people’s health. Severe air pollution can cause respiratory diseases, while good air quality is beneficial to physical and mental health. Therefore, the prediction of air quality is very important. Since the concentration data of air pollutants are time series, their time characteristics should be considered in their prediction. However, the traditional neural network for time series prediction is limited by its own structure, which makes it very easy for it to fall into a local optimum during the training process. The empirical mode decomposition fuzzy forecast model for air quality, which is based on the extreme learning machine, is proposed in this paper. Empirical mode decomposition can analyze the changing trend of air quality well and obtain the changing trend of air quality under different time scales. According to the changing trend under different time scales, the extreme learning machine is used for fast training, and the corresponding prediction value is obtained. The adaptive fuzzy inference system is used for fitting to obtain the final air quality prediction result. The experimental results show that our model improves the accuracy of both short-term and long-term prediction by about 30% compared to other models, which indicates the remarkable efficacy of our approach. The research of this paper can provide the government with accurate future air quality information, which can take corresponding control measures in a targeted manner.
BibTeX
@inproceedings{jiang2022empirical,
  author = {Wenxin Jiang and Guo-Chang Zhu and Yi-Yun Shen and Qian Xie and Min Ji and Yongtao Yu},
  title = {{An Empirical Mode Decomposition Fuzzy Forecast Model for Air Quality}},
  booktitle = {Entropy},
  year = {2022},
  doi = {10.3390/e24121803},
}
Figure