View Proposal

Proposer
Sarat Dass
Title
Kolmogorov–Arnold Network (KAN) for Ozone Prediction in Industrial Cities Using Meteorological and Pollutant Time Series Data
Goal
Develop a predictive model for ozone concentrations in industrial cities, supporting better urban air quality management and early warning systems.
Description
The purpose of this project is to develop a predictive model for ozone concentrations in industrial cities, supporting better urban air quality management and early warning systems. Due to the complex, non-linear interactions between meteorological conditions and pollutant levels, traditional statistical models often fall short in capturing these dynamics. Classical machine learning methods, such as decision trees or support vector machines, can improve prediction accuracy but may struggle to fully capture long-range temporal dependencies and subtle variable interactions in time series data. On the other hand, neural networks are well-suited for modelling such non-linear relationships and temporal patterns. While conventional deep learning models demonstrate strong predictive performance in environmental forecasting, their black-box nature limits interpretability, which is a key requirement for understanding environmental processes. Kolmogorov–Arnold Networks (KAN) offer a promising alternative by combining the flexibility of neural networks with the transparency of functional decomposition. Given the temporal structure of the dataset, KAN offers a compelling framework to explore both predictive accuracy and interpretability in time series ozone forecasting.
Resources
1. Gao, Y., Hu, Z., Chen, W.-A., Liu, M. and Ruan, Y. (2025). A revolutionary neural network architecture with interpretability and flexibility based on Kolmogorov–Arnold for solar radiation and temperature forecasting. Applied Energy, 378, p.124844. doi:https://doi.org/10.1016/j.apenergy.2024.124844. 2. Wang, L., Zhao, Y., Shi, J., Ma, J., Liu, X., Han, D., Gao, H. and Huang, T. (2022). Predicting ozone formation in petrochemical industrialized Lanzhou city by interpretable ensemble machine learning. Environmental Pollution, [online] 318, p.120798. doi:https://doi.org/10.1016/j.envpol.2022.120798
Background
Url
External Link
Difficulty Level
Moderate
Ethical Approval
None
Number Of Students
1
Supervisor
Sarat Dass
Keywords
data analysis, time series, machine learning, kan networks, interpretable machine learning.
Degrees
Bachelor of Science in Statistical Data Science
Back to List