Hosting Customer Clustering Based On Log Web Server Using K-Means Algorithm

Abstract

To compete in global industries, a company must have a good business strategy. Especially for domain and hosting company that has many competitors there. The business strategy could be found with hosting customer behavior based on log web server analytics. The most important log web server associated with customer access is recorded in the access.log file. Potential customers were read from access activity in the form of request method /pesan on access.log. One of popular method for data mining from log server is Clustering with K-Means Algorithm. This algorithm was chosen because K-Means has a fast execution time, easy to implement, and good for a big numeric data. The evaluation technique determining the optimal value of K is used Elbow Method and the Calinski Harabasz Index. K-Means algorithm can be used to determine the pattern of hosting customers based log web server. The results of this research indicate that the clustering process based on web server log with K-Means Algorithm can be used to know the pattern of customer hosting. There are total 5 clusters for data by week and data access time. The pattern of hosting customers that are formed in ordering a succession of clusters 1,2,3,4,0. The most ordered areas are Jakarta in cluster 1, Bandung Semarang, Surabaya on cluster 2 and Medan, Tangerang, Malang, Yogyakarta on cluster 3. The frequency of booking is mostly done at the beginning of the month at 12.00 - 23.59. This customer behavior could be a reference to know the best business strategy to expand the marketing in cluster 4 and 0 and help any other stakeholder for making some policy to develop the company.