Predicting-Air-Quality

Team Members:

Varsha Suresh (015245347)

Lasya Bheemendra Nalini (014618695)

Auni Bagchi (011256895)

Project Title: Predicting Air Quality and Pollutant Levels in the US

Project Report: https://github.com/lasyabheemendra/Predicting-Air-Quality/blob/main/FA21%20CMPE%20255%20Term%20Project%20Report%20-%20Group%209.docx

This directory contains the work done by all team members of Group 9 towards the term project. This document can be used as a README file and will describe the different files present here along with the purposes they serve.

Dataset: The dataset has been divided into 5 sections, each of which is devoted to a pollutant. The daily summary data of each pollutant for the years 2000-2020 has been uploaded to the following five sub directories. They collectively form the dataset for the project.

1. CO
2. SO2/data
3. NO2
4. Ozone
5. PM2.5/data

US Census data for population estimates: FA21 CMPE 255 Term Project/co-est2020.csv EPA site data for land use and location setting: FA21 CMPE 255 Term Project/aqs_sites.csv

Preprocessing and prediction code Files: Within each of the aforementioned subdirectories, colab files/ipynb files are present which contain the source code used to preprocess the data and to generate the AQI for test data(years 2018-2020). File paths are as follows:

1. CO/CO_GradientBoost.ipynb
2. NO2/NO2LinearRegression.ipynb
3. Ozone/OzoneGradientBoost.ipynb
4. PM2.5/PM2.5GradientBooster.ipynb.ipynb
5. SO2/SO2_GradientBooster_AQI.ipynb

Prediction files for test data: Within each of the pollutant subdirectories, the AQI prediction files are also contained.

1. SO2/SO2_AQI_Predicted
2. PM2.5/PM2.5_AQI_Predicted
3. Ozone/Ozone_AQI_Predicted
4. NO2/NO2_AQI_Predicted
5. CO/CO_AQI_Predicted

AQI prediction processing files:

The following files have been used to explore the predicted AQI for all five pollutants. The data is processed differently in each file and has been used iteratively to build the visualizations for this project. The files and their processing steps is detailed below:

1. Combined Pollutants/PollutantMerge_ver1.ipynb Initial analysis to generate visualizations. The mean of AQI per month was calculated and the resulting dataset was merged with population and site information to draw up graphs. These graphs were used to analyze associations between the different datasets. Result file: Combined Pollutants/PredictedAQIData.csv Associated visualization (not most recent): FA21 CMPE 255 Term Project/Visualizations/AQI_Visualization_Subset.twbx

2. Combined Pollutants/PollutantMerge_final.ipynb Modified version of PollutantMerge file, to address bugs in processing which did not generate accurate visualizations. In this file, the mean of AQI over a month is not calculated. Result file: Combined Pollutants/predictedQualityWithSite1.csv Associated visualization: Visualizations/AQIDays1.twbx

Association Analysis: Association Analysis has been performed and the code files for train and test data can be found as detailed below: Train data: Association Analysis/AssociationAnalysis_TrainData.ipynb Test data: Association Analysis/AssociationAnalysis_predictedData.ipynb

Comparing different Algorithms: Before selecting Gradient Boosting Regressor, few other algorithms were tested for PM2.5 and SO2 pollutants. Below files show the results obtained from it.

1. PM2.5/PM2.5_algorithms_AQI.ipynb
2. SO2/SO2_algorithms_AQI.ipynb

The comparison of different algorithms for other pollutants was included within these files:

1. CO/CO_GradientBoost.ipynb
2. NO2/NO2LinearRegression.ipynb
3. Ozone/OzoneGradientBoost.ipynb