TitleSpatialization of Web Sites Using a Weighted Frequency Model of Navigation Data
Publication TypeJournal Articles
Year of Publication2004
AuthorsReitsma, R, Thabane, L, MacLeod, J
JournalJournal of the Association for Information Science and Technology
Date Published2004

Reitsma, Thabane, and MacLeod are interested in the display of document sets as visualized geometric spaces. Such spaces can use metrics and dimensions determined arbitrarily prior to analysis of data, or they may use secondary data (logged website transaction counts, perhaps) with techniques like factor analysis or MDS to find a structure. Using high transaction volume between an origin and a destination as an indicator of a small distance and a low volume as an indicator of a large distance, a transaction log can provide input to MDS. One problem is the possible origination of multiple sessions from the same address where one can not determine if consecutive requests are part of the same transaction and thus frequencies may be invalid. They suggest the use of the probability that a count is a transaction as a weight rather the count alone, with this probability depending upon the time separation between an origin and a destination with less time indicating a higher probability. A transaction log for a website for undergraduate engineering learning was analyzed in this manner and weighted transaction counts were compared to the use of straight count inputs to MDS using the Euclidean metric and four dimensions. Weighted results were not significantly different.