Development of a BP Neural Network-Based Model for Prediction of Thermal and Humidity Parameters and Particulate Matter Concentration in Open Space

This paper takes the monitoring data of 14 open spaces in Nangang District as the research object, selects the main influencing factors of thermal-particulate matter as the input variables, and uses BP neural network to realise the thermal-particulate matter prediction in hourly scale. Doing a good job of predicting the thermal environment and particulate matter concentration is beneficial for pedestrians to prepare for travelling, which can effectively protect their health. The traditional multiple linear regression prediction model has more limitations and the accuracy of the prediction results is low. In this paper, neural networks are chosen to make predictions. Artificial neural networks have a strong learning ability, can reflect the non-linear influence of the role of influencing factors, establish complex non-linear relationships, and are better able to grasp the intrinsic laws of the data, avoid the shortcomings of multiple linear regression models, and improve the accuracy of prediction.