Skip to content

In this project, we develop a machine-learning model for air-quality prediction in Poland

License

Notifications You must be signed in to change notification settings

OmdenaAI/warsaw-poland-chapter-air-pollution

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Mitigating Air Pollution in Poland Through Machine Learning

Project summary

The aim of this project was to investigate the primary factors responsible for air pollution in Poland and develop an effective tool to predict air quality. To achieve this, three datasets were utilized: daily air quality data, daily weather data, and Static annual data from the Polish Central Statistical Office, all from 2017 to 2021. The annual data provided information on various features related to human acitivities, such as area by land use, crop production, emission of particulates and pollutant gases, forest area and fires, population density, production of electricity, vehicle types, and air pollution reduction systems. To achieve the project's objectives, a methodology was adopted, which involved translating and cleaning the data, combining data from different years and pollutants, followed by data analysis and machine learning. The cleaned datasets were analyzed to identify the key factors contributing to air pollution in Poland.

The results revealed that the concentration of pollutants varies during different times of the year. PM10 and PM2.5 are primarily emitted from sources such as solid fuel combustion, road traffic, dust from construction sites, and agricultural activities. Industrial plants such as coking plants, steel mills, and coal-fired power plants are also significant sources of air pollution in many places in Poland. NO2 is primarily emitted from combustion processes in energy production, manufacturing industry, and road transport, with diesel engines being a significant contributor to nitrogen oxide emissions. O3 is a secondary pollutant formed by the influence of solar radiation on a combination of airborne pollutants, including NOx and VOCs.

To standardize the measurement of air quality and communicate it to the public, the Common Air Quality Index (CAQI) was used. It considers various pollutants, including PM10, PM2.5, O$_3$, NO$_2$, and SO$_2$, and assigns a value to indicate the level of air pollution. The CAQI levels are higher during weekdays and lower during weekends, potentially indicating that weekly pollutant concentrations depend on human activities.

The dataset was preprocessed, and two datasets were created for training classification and regression models. Various techniques, including feature engineering, feature selection, cross-validation, and hyperparameter optimization, were utilized. A range of models were trained and tested to predict CAQI classes and indexes, and their performance was evaluated using metrics such as F1 score and RMSE. Finally, an XGBoost regressor model was used to forecast future CAQI levels of Warsaw, and the model’s short-term forecasts was really close to the actual CAQI levels for the forecasting period.

Daily monitoring and recording of the various features that affect air quality in Poland can play a significant role in identifying the overall air quality levels. By analyzing and interpreting such data, policymakers and stakeholders can gain valuable insights into the trends and patterns of air pollution in the region. This information can help them identify areas of concern and take appropriate actions to mitigate the impact of air pollution on public health and the environment. Overall, a data-driven approach to air quality monitoring and policymaking can contribute significantly to creating a healthier and sustainable environment in Poland.

More details of this project can be found in the project report inside the report folder.

Data Sources

We used publicly available data sources:

Task overview

This project was divided into different tasks:

  • Task 1 - Data Cleaning:

    • Cleaning data files from the data sources
    • Obtaining combined tabular datasets for (i) pollutant data, (ii) weather data, and (iii) general statistical data
  • Task 2 - Data Preprocessing:

    • Merging locations of measurement stations for pollutants and weather appropriately
    • Combining them with the general statistical data
    • Obtaining a single tabular data file that can be used for EDA and modelling
  • Task 3 - EDA:

    • Filling missing values
    • Data visualization and time-series analysis
    • Pinning down main causes and possible hotspots of air pollution
  • Task 4 - Modelling:

    • Developing a machine-learning model to predict air quality based on the given features
    • Both classification models (air quality categories) and a regression models
    • Using the trained models to analyze feature importance to identify main factors contributing to air pollution in Poland

Repository Structure

├── LICENSE
├── README.md          <- The top-level README of this project.
│ 
│
├── report             <- Folder containing the final report of this project
│ 
│   
├── src                <- Source code folder for this project
    │
    ├── data           <- Datasets used in this project
    │   
    ├── tasks          <- Master folder for all individual task folders
    │
    └── visualizations <- Code and Visualization dashboards generated for the project

About

In this project, we develop a machine-learning model for air-quality prediction in Poland

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published