Visual Analysis of E-Commerce User Behavior Based on Log Mining

Table 1

Comparison between K-means and other clustering algorithms.

Algorithm	Feature	Disadvantage	Description
CLIQUE	The speed is independent of the number of data objects and only depends on the number of cells in each dimension in the data space.	Parameter sensitive, unable to deal with irregularly distributed data, dimension disaster, etc.	The CLIQUE algorithm cannot meet the requirements of e-commerce platform by exchanging efficiency for accuracy.
FCM	The clustering effect will be very good for the data satisfying the normal distribution, and the algorithm is sensitive to outliers.	FCM cannot be guaranteed to converge to an optimal solution, and the performance of the algorithm depends on the initial clustering center.	The K-means algorithm has fast clustering speed and good clustering effect.
DBSCAN	Irregularly shaped clusters can be resolved, and it handles noisy data well.	For clusters with different densities, the DBSCAN algorithm may not work very well.	The parameter selection of DBSCAN algorithm requires manual intervention, which is slower than that of K-means.