Air pollution is garnering significant attention as a possible hazard to human health. As a result, ensuring adequate air quality performance becomes a pressing public health priority. Governments and communities are getting more worried about air pollution, which has detrimental effects on global human health and sustainable development. Today, the most popular technique for estimating air quality is CNN. However, these methodologies provide poor results, compelling us to predict air quality using deep architecture models. This study used the ARIMA model and regression technique to construct a deep learning approach for predicting air quality categorization (AQC) in big data. For final categorization, the Linear Regression approach was used. Using a Deep Neural Network, a substantial prediction model is created. DNN can analyze and remember consecutive data over time, such as daily data on air quality. Unlike typical time series prediction models, this model can predict air quality at every station at the same time and exhibit seasonal stability. The supplied models outperformed the suggested strategy in forecasting air quality. This research identified PM 2.5 characteristics for using real-time sensors with IoT. People are becoming more aware of the amount of air pollution in areas such as hospitals, schools, and other public spaces to predict the level of PM2.5 pollution.