Collaborative filtering based on demographic attribute vector

In present recommender systems, users receive items recommended on basis of their purchase records. New user experiences the cold start problem: as there records is very poorly. This paper proposed an NCT/TF(number of common terms / term frequency) collaborate filtering Algorithm Based on demographic vector. First, generates user demographic vector base on the user information (age, occupation, gender). then calculate two users similarity base on previous result. and generate new similar by combine it with cosine or PCC similar And then predict item rates by top N similar neighbors. The experiments show that the quality of recommendations improved, while the new user effort is smaller as no initial rating are asked.