Analysis of classification and clustering techniques for ambient AQI using machine learning algorithms

The ambient air quality index impacts people's health. The air quality index of any place is measured and analyzed in everyday routine. The AQI information helps the people to know and causes of the major pollutants. The objective of the proposed study on the classifications and cluster of ambient air quality index of south Indian cities This dataset about the pollutants information for analysis is considered the last six years' data from 2015 to 2020. Various data preprocessing techniques are applied on datasets for handling missing values and range normalization used on the features. Feature selection used the Person Correlation with the ranker search algorithm to evaluate the worth of attributes and the class variable. A detailed analysis was performed using various classification algorithms such as the Naive Bayesian model and Bayes network model, the regression models such as support vector regression, logistic regression, and multi-layer perceptron model, the instance-based models such as KNN and locally weighted regression, the ensemble models such as Bagging and Boosting model and Tree-based models such as J48 and Random forest are used. From the clustering technique, k-means clustering and EM clustering were used for the analysis. Results display a comparative analysis on all the experimented techniques based on the various terms such as accuracy, RMSE, and time to build the model. WEKA-3.8.5 Tool is used for preprocessing and implementation of the machine learning models and the accuracy comparison is discussed.