ABSTRACT
In order to prevent the impact of sparsity, the paper proposes a new calculation method of similarity: similarity calculation method based on principal component. Principal component analysis is a method of mathematical transformation, which the given set of relevant variables turn into another set of uncorrelated variables (principal components) by a linear transformation, these new variables (principal components)
are in descending order according to the variance order [7,8,9]. If the user (item) rating data sees it as a set of correlated variables, then through the principal component analysis it can get the group of unrelated user (item) principal components. Similarity calculation on the principal component not only can avoid the influence of the data sparsity but also can reach the purpose of reducing the dimensions.
