Web User Action Clustering Algorithm
Ning Ye · Mini-micro Systems · 2004
A clustering algorithm called FCC (Filter Coefficient Clustering) is proposed in this paper to get similitude action of Web users through mining web logs. First, we define a novel similarity coefficient called CM, and then combine it with the Jaccard Coefficient. As a result, a mix Coefficient is obtained to represent the similarity of Web user actions. The main reason why our algorithm is superior to other algorithms such as hierarchical clustering is that a threshold is engaged to filter the web log, thus the volume of data processed declined dramatically and the dimension disaster problem is resolved. The experiment result is satisfactory.