ikoloska / pm2.5

Predicts the level of PM2.5 particles in one of the most polluted cities in Europe you've never heard of

PM2.5

Predicts the level of PM2.5 particles in one of the most polluted cities in Europe.

PM2.5 Revisited

Skopje, the capital of the Republic of Macedonia, is one of the most poluted cities in Europe. This project, predicts the level of PM2.5 particles which can enter the bloodstream and cause heart and lungs issues, by using the level of Carbon Dioxide (CO2), Nitrogen Dioxide (NO2), Trioxygen/Ozone (O3) and PM10 particles.

Prerequisites

The code requires:

• Numpy

$ pip install numpy

• Pandas

$ pip install pandas

• Sklearn

$ pip install sklearn

Dataset

The data set can be found on https://www.kaggle.com/cokastefan/pm10-pollution-data-in-skopje-from-2008-to-2018. However, it is quite sparse and requires some preprocessing. The measurements are aggregated based on geographical location (I've included an example). The PM2.5 particles are then divided into 20 equaly spaced categories.

Results

The highest overall prediction accuracy is around 80%, by using a Random Forest Classifier with 100 estimators. More details on RFC can be found on http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html https://medium.com/machine-learning-101/chapter-5-random-forest-classifier-56dc7425c3e1 https://en.wikipedia.org/wiki/Random_forest

Further reading

Check out https://www.dw.com/en/skopje-welcome-to-europes-most-polluted-city/g-42083092 for info on Skopje's polution.

Ackgnoligments

Thanks to cokastefan (https://www.kaggle.com/cokastefan) for uploading the dataset to Kaggle.