Table 1.

Research gap

Related Work	Difference
Idowu and Kattukottai [9]	The related work clustered purchase behaviour whilst this paper clustered the underlying visit intents. Whilst purchase clusters are important, an intent model is equally important as not every visitor would commit to a purchase and not all websites have purchase applications
Porsche et al. [10]	The related work employed custom metrics and descriptive statistics to track visitors reading behaviour. However, this paper employs clustering models which is an advanced analytics technique that would allow better generalization of the underlying intents as opposed to descriptive statistics
Domazet and Simovic [11]	The related work employed descriptive statistics to measure a website’s performance. However, this paper employs clustering models which is an advanced analytics technique that would allow better generalization of the underlying intents as opposed to descriptive statistics
Jonathan et al. [12]	Within the related work, the author does not discuss much about the analytics employed. Nonetheless, the audience behaviour expected on a church website would fundamentally differ from the studied website within this paper
Semeradova and Weinlich [13]	The related work here proposed web analytics using the Google Analytics interface which was at an aggregate level. Aggregate analytics will provide a high-level overview which assumes that every visitor has behaved in this manner. The methods employed within this paper allow for better visitor profiling
Kalyankar and Anute [14]	The related work did not employ clustering methods to understand the visitors’ behaviour. However, this paper employs clustering models which is an advanced analytics technique that would allow better generalization of the underlying intents as opposed to descriptive statistics
Cirlugea et al. [15]	The related work focused on the effect of marketing on website volumes. However, did not discuss the quality of the visit by detailing the visitors’ activities once on the website
Rosqa and Ati [16]	The related work here was primarily exploratory
Pirvu and Anghel [17]	The aim of the related work here was to predict behaviour but the authors did not explain the different types of behaviours or events being predicted. A clustering method (as proposed within this paper) should have been first determined
Mariyapillai and Pratheepan [18]	The related work here was primarily exploratory to determine the geo-location of the visits
Stelian and Stoicu-Tivadar [19]	The related work here was primarily exploratory with main interest in visitor interaction with virtual bone structures. Clustering may not be appropriate within this application if a visitor was guided through the application as opposed to a website where a visitor is free to behave as desired