An Improved Condensing Algorithm

Abstract

kNN classifier is widely used in text categorization, however, kNN has the large computational and store requirements, and its performance also suffers from uneven distribution of training data. Usually, condensing technique is resorted to reducing the noises of training data and decreasing the cost of time and space. Traditional condensing technique picks up samples in a random manner when initialization. Though random sampling is one means to reduce outliers, the extremely stochastic may lead to bad performance sometimes, that is, advantages of sampling may be suppressed. To avoid such a misfortune, we propose a variation of traditional condensing technique. Experiment results illustrate this strategy can solve above problems effectively.

A concept drift-tolerant case-base editing technique

“…Although intuitive, CNN has been criticized for being sensitive to the order of cases examined and to noise [5]. To overcome these problems, several modifications to CNN have been made, such as the Reduced Nearest Neighbor (RNN) [42], the Selective Nearest Neighbor (SNN) [43], the Modified CNN (MCNN) [44], the Generalized CNN (GCNN) [45], the Fast CNN (FCNN) [6] and the Improved CNN (ICNN) [46]. Other approaches to case editing build a competence model of the training data and use the competence properties of the cases to determine which cases to include in the edited set.…”

Adaptive case-based reasoning using retention and forgetting strategies

“…Next, some improvements over the CNN rule were presented: RNN (Reduced Nearest Neighbour rule) [51] and SNN (Selective Nearest Neighbour) [52]. Furthermore, in recent years, different variations of the CNN rule have also been proposed: GCNN (Generalized Condensed Nearest Neighbour Rule) [53]; FCNN (Fast Condensed Nearest Neighbor rule) [54]; and a variation for text categorization [55].…”

The Good, the Bad and the Incorrectly Classified: Profiling Cases for Case-Base Editing

“…This technique will allow for the removal of noisy cases but is sensitive to the order of presentation of cases. More recent improvements to CNN have been proposed by Chou et al [8] and Angiulli [9] with Hao et al [10] proposing a variation appropriate for text classification.…”