Introduction to statistical pattern recognition (2nd ed.)
Introduction to statistical pattern recognition (2nd ed.)
Pattern recognition: statistical, structural and neural approaches
Pattern recognition: statistical, structural and neural approaches
Pattern Classification (2nd Edition)
Pattern Classification (2nd Edition)
Nearest neighbor pattern classification
IEEE Transactions on Information Theory
Margin calibration in SVM class-imbalanced learning
Neurocomputing
On-line evolving image classifiers and their application to surface inspection
Image and Vision Computing
| Hi-index | 0.00 |
In this study-aimed to achieve optimal accuracy in the classification of rotifers according to the number of eggs carried-several modifications to the basic kNN method have been proposed and assessed. Six distinct kNN rules as well as several additional hybrid models were, in fact, devised or employed and their precision compared. Meanwhile, the data sets used in the evaluation of each of these methods were acquired from rotifer images generated via the shape moments approach. Both the original data sets and the edited ones, formed by removing outliers from the originals, were used in the evaluation of these adjusted models. Through a process of comparative evaluation, several of the modified algorithms proposed-comprising both individual and hybrid models-were found to perform better overall than the classical kNN method. Refinements related to class-size weighting, in particular, were shown to heighten the accuracy of the classical kNN model considerably. Close evaluation of the various models created revealed kNN-CCS and F-kNN-CCS, in their application to the edited data sets, to be the most reliable individual modified and hybrid models respectively, with levels of accuracy greater than 95%.