Contemporary Classification on Medical Data based on Non-Linear Feature Extraction
Thannob Aribarg, Siriporn Supratid, Chidchanok Lursinsap · 2009
High dimensional data in several applications seriously spoils classification computation of several types of learning. In order to relieve the difficulties of such a high dimension, this paper proposes the classification computation, which refers to a modified neural network: the neural network with weights optimized by particle swarm intelligence. The contemporary is placed on the combination of the non-linear feature extraction and such a classification method. 10-fold cross-validation experiments of each method are performed on five medical data sets. The results indicate not only the improvement of classification based on non-linear feature extraction, but also indicate the reduction of the number of features for classification.