Abstract–The quantity of text information published inArabic language on the net requires the implementation ofeffective techniques for the extraction and classifying of relevantinformation contained in large corpus of texts. In this paper wepresented an implementation of an enhanced k-NN Arabic textclassifier. We apply the traditional k-NN and Naive Bayes fromWeka Toolkit for comparison purpose. Our proposed modifiedk-NN algorithm features an improved decision rule to skip theclasses that are less similar and identify the right class from knearest neighbours which increases the accuracy. The studyevaluates the improved decision rule technique using thestandard of recall, precision and f-measure as the basis ofcomparison. We concluded that the effectiveness of the proposedclassifier is promising and outperforms the classical k-NNclassifier.
展开▼
