Abstract: Basically, the PageRank algorithms set the inputted items with initial ranks. They usually use a normal distribution of initial rank in the items space with the ratio of 1/n for each item where n is the number of items in the dataset. Alternatively, a dynamic distribution of initial ranks according to the user preferences and items features has been proposed for recommendation system improving. The adapted PageRank algorithm was evaluated in comparis on with the traditional algorithm. The results were extremely encouraging with respect to the recommendation system improvement. The evaluation measures ofthe suggested algorithm were examined on the MovieLens dataset.