Image classification of malaria using hybrid algorithms: convolutional neural network and method to find appropriate K for K-nearest neighbor
10.11591/ijeecs.v16.i1.pp382-388
Wisit Lumchanow
,
Sakol Udomsiri
This paper presents image classification algorithms to improve the learning rate and to comparison the classification efficiency. Using convolutional neural network (CNN) for feature extraction and method to find appropriate k for k-nearest neighbor (KNN). Medical datasets were used in the experiments to classify Plasmodium Vivax and Plasmodium Falciparum. Results of the study indicated that for Plasmodium Vivax in ring form, the appropriate k was 1 and the learning rate (LR) was 83.33%, Trophozoite (k=5, LR=91.67%), Schizont (k=1, LR=83.33%), and Gametocyte (k=1, LR=91.67%) whereas Plasmodium Falciparum in ring form (k=7, LR=91.67%), Trophozoite (k=1, LR=83.33%), Schizont (k=1, LR=91.67%) and Gametocyte (k=1, LR=100%).
