In today’s world, people are more concerned about the quality of air. Since air is everywhere, we cannot escape from the pollutants of air. To keep our health safe, we need to maintain certain quality air in our surroundings. Effective air prediction is the most trending research in this era. This paper discusses some of the challenges that are faced due to lack of data resources, various concentrations of the quality of air, etc. And we propose a solution to these kinds of problems with the help of a predictive data feature extraction based on an approach of predicting the quality of air. This model is based on lightBGM which predicts PM2.5 concentration of air at 35 air monitoring systems at Beijing in the upcoming 24 hours. This high dimensionality large-scale data is collected and employed using the CNN, KNN, and random forest algorithm which is used to predict the quality of air. As you explore new data, you can use the spatial data remaining in the existing model. We can also improve the predictive accuracy of the existing model and make our model more efficient. Using the sliding window mechanism, we can mine deeper data of high dimensions, increasing the amount of information. Here we implement the comparison of the predicted values with the input actual values from the data set which was provided and prove that our model is more beneficial and superior to all other models that already exist by constructing a high-dimensional statistical analysis on the data the implementation give the effective results over the air quality management.