Exploitation of Neural Methods for Imputation

In this presentation I will discuss modern imputation methods based on the neural nets methodology. The most important method used here is the Tree-Structured Self-Organising Map, or TS-SOM. The TS-SOM is a computationally fast variation of the basic Self-Organising Maps, or SOMs. It is a combination of the SOM, tree-structured clustering and computational speedup techniques. SOM is an iterative method for classification and can thus also be used for finding homogeneous clusters suitable as multivariate imputation classes. MLP (Multi-Layer Perceptron) and SVM (Support Vector Machines) are considered briefly from the point of view of imputation. Along with many other modern methods, TS-SOM is included in a versatile software program entitled NDA, or Neural Data Analysis, which was created and will be maintained by a research group on Software Engineering and Computational Intelligence of the University of Jyväskylä, Finland. Imputation methods have been implemented into NDA in co-operation with a research group of Statistics Finland. This presentation is based on research conducted under the EUREDIT FP5 project of the European Union.