The air quality predication based on machine learning methods

Abstract

Air quality is the focus of attention all over the world. As an important index to measure air quality, accurate prediction of PM2.5 plays an essential role in regulating air quality. This paper uses different machine learning algorithms and neural networks to infer air quality (AQ) in the study area and observe the impact of these methods on accuracy. The results indicate that shuffling the data can enhance the model's performance. In addition, the neural network is the most affected by the data shuffle operation compared to other models. In the case of a shuffle operation, the performance of the neural networks is the lowest among all models. However, in the case of non-shuffle data, the neural network performance is the best among all models. Therefore, in the absence of large-scale data sets, the traditional machine learning method with a relatively small scale should be selected to model the air quality prediction problem because the traditional machine learning method performs better in the small sample data scene.