Probabilistic Model Checking of DTMC Models of User Activity Patterns

Abstract

Software developers cannot always anticipate how users will actually use their software as it may vary from user to user, and even from use to use for an individual user. In order to address questions raised by system developers and evaluators about software usage, we define new probabilistic models that characterise user behaviour, based on activity patterns inferred from actual logged user traces. We encode these new models in a probabilistic model checker and use probabilistic temporal logics to gain insight into software usage. We motivate and illustrate our approach by application to the logged user traces of an iOS app.

Probabilistic Formal Analysis of App Usage to Inform Redesign

“…We note that the distributions of the two activity patterns in the population of users are similar for the time intervals [0, 1) and [30, 60) -probably because more users exhibit a more exploratory behaviour during these times (new types of usage statistics become available after one month of usage). At the same time, the plots for the time intervals [1,7), [7,30), and [60, 90) are also similar, and we think that they correspond to a settled (or routine) usage behaviour.…”

“…We first introduced the concept of representing the behaviour of users through a weighted mixture over data gen-erating distributions [11], refining the concept substantially in [1] where we defined activity patterns for an individual user as user meta models (DTMCs), with respect to a population of users. We then inferred behaviours for individual user activity from large scale logged usage data for a mobile game app and analysed them using probabilistic temporal properties (without rewards).…”

A probabilistic approach to user mobility prediction for wireless services

“…This does reinforce that there is a tangible need for PDF-predictors for use by services. Andrei et al [17] used an ExpectationMaximisation [14] (EM) algorithm to generate DTMCs that classify user activity patterns for their iOS app, Hungry Yoshi. The approach taken is very similar to what we use to generate our HEM predictor.…”

On Lions, Impala, and Bigraphs

“…Probabilities for these kinds of user-initiated events could be inferred from logs of user trials. For example, the recent work on inferring activity patterns from user logs [Andrei et al 2014] may indicate a fruitful direction. Here, a finite number of (activity) patterns of usage behaviour (sets of probabilities of transitions between states) are inferred from mass trials (e.g.…”