In this paper, a new approach for pre-processing stage of the web usage mining is proposed. The proposed method provides more successful results when dealing with raw log files. Moreover the complexity of the proposed method is considerably low. The related areas in the log files are determined for mining stage. Then, by using the association rules method the interesting patterns are extracted for consequent goals. In the result of this study, knowledge extraction was performed by proposed method. The significant and pure knowledge is contributed to structural organization of website. Besides, web site development and about the availability of the site to give advice to managers and designers are available.