Effects of term weighting approach with and without stop words removing on Arabic text classification

Text classification is a technique for grouping articles into pre-defined categories. Prior to categorization, text documents have to be set up and organised in a form that is compatible with the data mining techniques. That has produced many term weighting techniques have been developed in the literature to improve the functionality of text categorization systems. This paper compares techniques for text classification based on when stop words are eliminated from the documents once and when they are not. We used an Arabic data set of 322 papers grouped into six subjects (health, politics, science, education, agriculture and finance) to research which is the influence of prior weighting of features approaches on classification outcomes with regard to accuracy, recall, precision, and F-measure values. Each document subset includes 50 documents, except the health one that includes 61. We carried out four experiments and checked results getting that only for precision, the term frequency feature weighting approach with stop word removal are more or less the same than for binary method. Moreover the conclusion for accuracy, recall and F-measure is that Feature weighting approach with stop word outperforms the Binary one. Additionally, if the phrase for weighting is the same for both methods, to remove stop word increases classification accuracy for all the experiments.