Automating Software Customization via Crowdsourcing using Association Rule Mining and Markov Decision Processes

As systems grow in size and complexity so do their configuration possibilities. Users of modern systems are easy to be confused and overwhelmed by the amount of choices they need to make in order to fit their systems to their exact needs. In this thesis, we propose a technique to select what information to elicit from the user so that the system can recommend the maximum number of personalized configuration items. Our method is based on constructing configuration elicitation dialogs through utilizing crowd wisdom. A set of configuration preferences in form of association rules is first mined from a crowd configuration data set. Possible configuration elicitation dialogs are then modeled through a Markov Decision Processes (MDPs). Within the model, association rules are used to automatically infer configuration decisions based on knowledge already elicited earlier in the dialog. This way, an MDP solver can search for elicitation strategies which maximize the expected amount of automated decisions, reducing thereby elicitation effort and increasing user confidence of the result. We conclude by reporting results of a case study in which this method is applied to the privacy configuration of Facebook.