Title :
A Bayesian approach to the missing features problem in classification
Author :
Lynch, Robert S., Jr. ; Willett, Peter K.
Author_Institution :
Naval Underwater Syst. Center, Newport, RI, USA
fDate :
6/21/1905 12:00:00 AM
Abstract :
In this paper, the Bayesian data reduction algorithm (BDRA) is extended to classify discrete test observations given the training data contains feature vectors which are missing values. Two methods are used to model missing features in the BDRA, where performance is compared to a neural network using both simulated and real data. In general, it is shown that the BDRA is superior to the neural network
Keywords :
Bayes methods; data reduction; pattern classification; BDRA; Bayesian data reduction algorithm; classification; discrete test observation classification; feature vectors; missing features problem; neural network; Bayesian methods; Contracts; Frequency; Hafnium; Neural networks; Probability distribution; Quantization; Random variables; Testing; Training data;
Conference_Titel :
Decision and Control, 1999. Proceedings of the 38th IEEE Conference on
Conference_Location :
Phoenix, AZ
Print_ISBN :
0-7803-5250-5
DOI :
10.1109/CDC.1999.827922
