There are disadvantages of traditional vector space model in computational complexity, query efficiency and intelligence. Based on the traditional vector space model, an intelligent N - dimension vector space model (INDVSM) is proposed. INDVSM shows distinct improvement in the computational complexity of document vectors by using a refined algorithm based on local information, and it takes the user's clicks as feedback to tune the initial vectors of documents and queries. Meanwhile the tuning algorithm is also presented and these two models are compared theoretically.