Over the past few years, more and more information has become available on the Internet. That is why it has become important to assist users' navigations by making personalized recommendations. In this paper, we propose to improve web recommendation by using a sequence-based model that allows noise during navigation. This characteristic is obtained by exploiting an enhancement of Markov models called skipping. The skipping based model we propose here runs using both a low time and space complexity and provides a high coverage. We focus on the importance to give to distant resources to increase the recommendation accuracy. In order to find the adequate importance of distant resources, we apply weighting schemes, and use the EM algorithm to look for the ideal weightings. The algorithm has been tested on an Intranet browsing dataset provided by a French bank. Results show that our model outperforms state of the art configurations, and constitutes a more tractable model. Indeed, the number of resources to take into account in order to improve the recommendation accuracy shows asymptotic above a low value. Last, the simple weighting scheme we propose provides results comparable to the results computed using the EM algorithm.