Design and implementation of network advertising precise marketing system based on parallel K-means algorithm

The research goal of this article is to find the actual distribution characteristics of the audience, that is that the audience was divided, and eventually find the factors affecting the distribution of the audience. Clustering is an important part in the whole research steps, shouldering on the responsibility of determining final classification success. If the clustering of the steps in the formation of the new distribution of audience is not accurate, so the following analysis of the influence factors are more impossible. Before clustering, it needs to determine the evaluation in the middle of the data clustering effect index, which can be an existing one or several attributes or by a few calculated a new attribute. What´s more, the original data can also be introduced according to need a determining factor.