Abstract:
App reviews and tweets from Twitter are currently being studied as the main source of information for analysing diferent aspects of software development. However, software related information does not only occur on these two platforms. Software forums are another source of software related information, with popular forums having millions of registered users and an active community. This thesis examines the content of software forums to determine whether or not software forums contain information to support the development of software. The data composition found in forums is compared to Play Store review data, and it was discovered that software forums contain more than double the amount of relevant information compared to Play Store review data. Using existing tools built for Play Store data to classify software forum data led to disappointing results. A new classifcation process is developed in this thesis to examine the feasibility of automatically classifying software forum data. The result showed that 74% recall and precision value can be achieved by using Metadata features and text features. The result of this study impact the design of potential new classifcation tools built to collect software requirements from software forums.
