Comparison of missing data filling methods in bridge health monitoring system

Abstract

In terms of the data characteristics of small sample, nonlinearity and seasonal regression in bridge health monitoring system, this paper analyses the applied results with different data filling methods such as linear regression, seasonal autoregressive integrated moving average (SARIMA), neural network BP approach and support vector machine (SVM). The comparison results show that support vector machines (SVM) and BP neural network have higher precision in the case of the same sample. The filling results show that support vector machines (SVM) has a higher accuracy than neural network BP with the small samples.