Comparison of Hot-Deck and Neural-Network Imputation

This chapter reports on an investigation to assess the ability of a neural network model to impute values in a travel survey by comparing its performance to that of a hot deck procedure and to the known values of the imputed variables. The possible alternative method used was an artificial neural network (ANN), which is often used in a wide variety of applications where responses can be learned. The analysis was directed at obtaining an assessment of both the absolute and relative performance of the ability of a neural network to impute values for variables in a typical travel survey. An empirical analysis was conducted using a sample drawn from the 1995 Nationwide Personal Transportation Survey in an effort to cast the investigation in the context of a typical, current travel survey. The analysis revealed that ANNs provide a viable alternative method of data imputation for travel surveys but the method has not been demonstrated to be superior to the hot deck procedure in this study.