Text mining IS ONE OF THE MAIN TASKS IN WEB RESEARCH THAT AIMS AT CLASSIFICATION OR CLUSTERING AVAILABLE TextS IN THE WEB FOR DIFFERENT APPLICATIONS, SUCH AS NEWS ANALYSIS AND SOCIAL NETWORK ANALYSIS. SINCE A VERY LARGE AMOUNT OF TextUAL DATA IS AVAILABLE ON THE WEB, REDUCING THE DIMENSION OF DATA USING FEATURE EXTRACTION TECHNIQUES PLAYS AN IMPORTANT ROLE IN IMPROVING THE EFFICIENCY AND EFFECTIVENESS OF THE Text mining ALGORITHMS. VARIOUS TECHNIQUES HAVE BEEN PROPOSED IN MACHINE LEARNING TASKS THAT CAN ALSO BE APPLIED IN THE Text mining DOMAIN. IN THIS PAPER WE STUDY THE AVAILABLE TECHNIQUES AND COMPARE THEIR IMPACT ON IMPROVING PERSIAN Text CLASSIFICATION PERFORMANCE. OUR EXPERIMENTAL RESULTS ON HAMSHAHRI CORPUS SHOWS THAT USING AN APPROPRIATE FEATURE SELECTION TECHNIQUE CAN IMPROVE THE CLASSIFICATION F-MEASURE FROM 88.12% TO 93.07%.