Abstract: Outlier detection is an important research area in text mining, information retrieval, machine learning, and statistics as well as enhancing natural language processing paradigms due to the enormous numbers of new documents being utilized for various information retrieval systems. One of the most challenging problems in this context is addressing the text categorization problem with Arabic text documents. In this paper, we propose a new text categorization (TC) algorithm which classifies Arabic text documents using the local sparsity coefficient-mine algorithm (LSC-mine algorithm). The chosen algorithm is capable of detecting outlier points in a spatial space and clusters documents by computing the LSC ratio between the new document and the cluster’s documents, which indicates the outlier-ness of a certain point. Several experiments have been conducted to ensure the success of the developed algorithm. Keywords: Text Categorization, LSC-mine, Arabic Language Text Clustering, Outlier Detection Algorithm.