Neural Network Based Techniques for Estimating Missing Data in Databases

Three techniques based on the application of neural networks to the estimation of missing data in a medical database are introduced. The approximation unit is introduced as a basic missing data estimation mechanism and is extended to an agent–based system, which provides improved accuracy. A view of the estimation of missing data from a classification perspective is also presented. The neural networks designed implement a multi–layer perceptron architecture with weight decay. The optimisation of the regularisation coefficient using the evidence framework is shown to improve accuracy in the estimation. Analysis of the results show that all methods proposed are comparable providing the same level of accuracy while maintaining the statistical properties of the original data set. The agent-based system is found to provide a better accuracy on average. Results also suggest that the method of approximation units can be used to classify the missing data into the categories of MAR or MNAR.

[1]  R.J. Marks,et al.  Implicit learning in autoencoder novelty assessment , 2002, Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290).

[2]  H. Toutenburg Little, R.J.A. and D.B. Rubin:Statistical analysis with missing data , 1991 .

[3]  Martin Fodslette Meiller A Scaled Conjugate Gradient Algorithm for Fast Supervised Learning , 1993 .

[4]  Tshilidzi Marwala,et al.  The use of genetic algorithms and neural networks to approximate missing data in database , 2005, IEEE 3rd International Conference on Computational Cybernetics, 2005. ICCC 2005..

[5]  Heekuck Oh,et al.  Neural Networks for Pattern Recognition , 1993, Adv. Comput..

[6]  R.G. Harley,et al.  Continuous on-line identification of nonlinear plants in power systems with missing sensor measurements , 2005, Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005..

[7]  J. Schafer,et al.  Missing data: our view of the state of the art. , 2002, Psychological methods.