Cosine similarity function is one of the most popular similarity function for handling the web data in various applications such as recommender system, collaborative filtering algorithms, classification algorithms, etc. Cosine similarity function calculates the similarity between the items as an angle between the vectors representing the two items. The smaller angle is for more similar items. In this paper, a new cosine similarity function is proposed that incorporates the temporal dynamic nature of the web data when calculating the similarity and provide more accurate results.