Predicting particulate matter (PM2.5) air pollution levels in Almaty city using machine learning techniques
Issakhov A. Rysmambetov N. Abylkassymova A.
August 2025Springer Science and Business Media Deutschland GmbH
Modeling Earth Systems and Environment
2025#11Issue 4
Air pollution is one of the important problems of large cities today. This paper is devoted to the development of methods for predicting the concentration of fine particles PM2.5 in the atmosphere of Almaty. The main objective of the study is to evaluate the effectiveness of different neural architectures for predicting the concentration of PM2.5 in the air of Almaty. The study used data obtained from sensors installed at different points throughout the city. The paper focuses on recurrent neural networks and their modifications: LSTM (seq2vec), bidirectional LSTM (BiLSTM) and Seq2Seq for predicting the concentration of PM2.5. This allows us to compare the effectiveness of models depending on the window size. The results showed that LSTM is better at forecasting for 90 days, Seq2Seq—for 180 days, and BILSTM—for 365 days. The application of these models can improve the air quality monitoring and management system in cities.
BiLSTM architecture , Ground sensors , LSTM architecture , Machine learning methods , Pollution level , SEQ2SEQ architecture
Text of the article Перейти на текст статьи
Kazakh British Technical University, Almaty, Kazakhstan
Al-Farabi Kazakh National University, Almaty, Kazakhstan
International Information Technology University, Almaty, Kazakhstan
Kazakh British Technical University
Al-Farabi Kazakh National University
International Information Technology University
10 лет помогаем публиковать статьи Международный издатель
Книга Публикация научной статьи Волощук 2026 Book Publication of a scientific article 2026