"Investigating Data Models for Automatically Generating Tests for Web Applications"

This paper looks at several factors that affect the parameters of a web page, and how they could allow us to test better and more efficiently. The factors that were explored include parameter interactions, history, and user roles. Parameter interaction has to do with looking at probabilities of a set, rather than individual parameters. History is fairly self-explanatory and looks at if a user's previous actions can help predict where they will go next. Lastly, user roles has to with a user's specific role or permission in a website and how that would affect the parameter values. The idea is that these factors could help us predict what parameter values should be so that we can develop more accurate test suites.

For me, this paper led me to have a lot of questions. In order to use these factors to predict parameter values, we need to be able to quantify them. This seems like a challenge, since things like user role and history may not be easily converted into something computational. Additionally, how do you actually figure out in what way a factor might affect a parameter value? Since there are many options for the values, this seems difficult.

This paper, and my question about it, lead directly into what I am working on. I am trying to determine how we can combine two or more factors and use joint conditional probability to improve the parameter values even more. This paper was very helpful to read because it really got me thinking about the various interactions between factors and parameters values.