A prediction system in user interest degree to Web sites using the concept of moving averages

Now that many organizations have invested a tremendous amount of money and efforts to operate Web sites on the Internet, there is a strong demand to understand the effectiveness of such investments. In other words, one of the most frequent and important questions about their Web sites is “Will the current Web site management policy be effective enough to have more visitors come to our Web site?” In this paper, a system which predicts the degree of user interest in the future to Web sites is constructed. The degree of user interest to a Web site is defined to be the visit counts for the Web site in the system. With higher the visit counts, the related site is considered to be more interesting. However, the figures of the visit counts themselves cannot explain properly the degree of user interest in the future to the related Web sites (i.e., the effectiveness of the related Web sites). Therefore, the system also uses mechanisms which use the concept of the Moving Averages, which have been used frequently in the stock exchanges. In this paper, two prediction mechanisms are proposed and compared. The first mechanism uses the Golden Cross/the Dead Cross of the Moving Averages, while the second uses the changes of upward/downward direction of the Moving Averages. Experimental results show that the two prediction mechanisms proposed in this paper predict the degree of user interest in the future to the related Web sites very well in most cases. However, the first one is considered to be better than the second one in the sense that the second one is too much sensitive to the changes of visit counts.