TY  - JOUR
ID  - timm2002fuzzyclusteranalysis
T1  - Fuzzy Cluster Analysis of Partially Missing Data Sets
A1  - Timm, Heiko
A1  - Döring, Christian
A1  - Kruse, Rudolf
JA  - Second International Workshop on Hybrid Methods for Adaptive Systems I
Y1  - 2002
SP  - 426
EP  - 431
N2  - A common problem in data analysis are missing attribute values in datasets. The easiest way to handle
such datasets in fuzzy cluster analysis is to discard data with missing values. Since this complete case
approach may result in a loss of valuable information and reduced dataset size, we study how missing values
can be handled by modified fuzzy clustering methods. These approaches are based on iterated imputation
of missing values, available case estimation of the cluster parameters, and the introduction of a class specific
probability for missing values. Benchmark datasets with randomly deleted attribute values are used to
demonstrate the capability of the presented approaches. Our experiments show that the modified clustering
methods are superior to a complete case analysis.
ER  -