Study on the impact of data preprocessing on performance of neural network for indoor airflow prediction

Abstract

In this study, neural network (NN） is implemented to predict non-isothermal indoor airflow and temperature distributions. Various data preprocessing methods are utilized and results are compared to reveal the impact of data preprocessing on NN performance. The results show that for most cases, different preprocessing methods can lead to similar NN performances with a prediction error of less than 5% for the mean value. Without data preprocessing for output, error submergent is likely to occur, and the gradient descent algorithm may fail to reduce errors of variables with smaller orders of magnitude during the training process.