Comparative Analysis of Neural Network and Machine Learning Techniques for Air Quality Prediction

Abstract

Empirical and naive methods used for predicting air quality is not fairly accurate. A system for predicting air quality using Machine Learning (ML) technique is presented in this paper. Dataset is collected from Delhi Meteorological department of a particular region of the year 2013-2016 consisting of nine parameters including PM2.5. Data set is labelled and divided to training and testing data. It undergoes various stages of pre-processing, learning and evaluation. ML techniques namely, Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF) and Neural Network (NN) are investigated to predict the quality of air based on the level of PM2.5 in the dataset collected. Model is created using the training data and performance is evaluated using the test data. Performance evaluation measures used are precision, recall, F1 score and support. Experimental results show that LR and SVM with hyperparameter tuning models give a better accuracy of 88.12% and 87.56% respectively, compared to other techniques.