Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Long short-term memory neural network for air pollutant concentration predictions: Method development and evaluation.

Xiang Li1, Ling Peng2, Xiaojing Yao2

  • 1Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100094, China; University of Chinese Academy of Sciences, Beijing 100049, China.

Environmental Pollution (Barking, Essex : 1987)
|September 14, 2017
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Precision chemical engineering of dendrimers for nucleic acid delivery.

Nature reviews. Chemistry·2026
Same author

Algorithmic management and employee innovation: unveiling the mediating mechanisms of perceived fit, perceived control and job crafting.

Frontiers in psychology·2026
Same author

Antioxidant lipid nanoparticles enhance mRNA stability for regeneration therapy and gene editing.

Nature communications·2026
Same author

Scallion White and Ginger Extracts Alleviate Stress-Induced Muscle Quality Deterioration in Crucian Carp During Transportation.

Foods (Basel, Switzerland)·2026
Same author

Fluorinated Amphiphilic Dendrimer to Improve PET Imaging of Cancer.

Small (Weinheim an der Bergstrasse, Germany)·2026
Same author

Development of High-Affinity CHD1 Chromodomain Inhibitors.

Journal of medicinal chemistry·2026

A new extended long short-term memory neural network (LSTME) model improves air pollutant concentration forecasting by capturing spatiotemporal correlations. This advanced model enhances public health warnings for pollutants like PM2.5.

Area of Science:

  • Environmental Science
  • Data Science
  • Artificial Intelligence

Background:

  • Air pollutant concentration forecasting is crucial for public health protection.
  • Existing prediction methods struggle with long-term dependencies and spatial correlations.
  • Accurate forecasting of pollutants such as PM2.5 is essential.

Purpose of the Study:

  • To propose a novel extended long short-term memory neural network (LSTME) model for air pollutant concentration prediction.
  • To effectively model spatiotemporal correlations in air pollutant data.
  • To enhance prediction accuracy by incorporating meteorological and timestamp data.

Main Methods:

  • Developed an LSTME model integrating Long Short-Term Memory (LSTM) layers for feature extraction.
Keywords:
Air pollutant concentration predictionsLong short-term memory neural network (LSTM NN)Multiscale predictionRecurrent neural networkSpatiotemporal correlation

Related Experiment Videos

  • Incorporated auxiliary data (meteorological, timestamp) to improve model performance.
  • Validated the model using hourly PM2.5 concentration data from 12 stations in Beijing (2014-2016) and compared it against STDL, TDNN, ARMA, SVR, and traditional LSTM models.

    • Main Results:

    • The proposed LSTME model outperformed traditional statistical and LSTM models in air pollutant concentration prediction.
    • The inclusion of auxiliary data significantly improved the model's predictive performance.
    • Achieved a Mean Absolute Percentage Error (MAPE) of 11.93% for one-hour predictions and 31.47% for 13-24 hour predictions.

    Conclusions:

    • The LSTME model demonstrates superior capability in capturing spatiotemporal dependencies for air pollutant forecasting.
    • The model provides a robust framework for enhancing air quality monitoring and public health advisories.
    • The LSTME model offers accurate predictions across various time scales, proving its practical applicability.