A Hybrid Predictive Modelling of Environmental Pollutant Concentrations using Machine and Deep Learning Techniques

Background: Diminishing air quality leads to serious risks to human beings, animals and surrounding environment. The pollution level indicator is a measure, which measure the cleanliness of the air, and it provides information on the level of polluted air and its potential impact on human health. Objectives: To propose a hybrid model using automated learning systems for predicting the AQI using historical air quality data. Methods: This approach involves cleaning the raw data, and model training using different machine learning and deep learning algorithms such as linear regression, decision tree, random forest, and convolutional neural network. Findings: Among all models, the RF_CNN algorithm gives the best results with a MAE of 19.26, MSE of 620.67, RMSE of 24.92, and R2 of 0.56. Novelty: This system predicts AQI by using automated learning algorithms and predict which learning algorithm produce better results. The results show that the deep learning models outperform existing machine learning models in predicting the index with best performance. The proposed approach can be used by environmental and zoological informatics, government agencies and policymakers to make data driven decisions regarding polluted air control, wild life conservation, eco system and health monitoring. Keywords: Wild life conservation, Environmental pollution, Model evaluation, Deep Neural Networks, Decision making