Mobile applications quality evaluation according to ISO 9126 quality model using Naïve Bayes

In this study a tool to evaluate the quality of mobile applications from the user provided reviews in Google Play app store is developed. ISO 9126 quality in use part is used as the quality model. We used a supervised Machine learning technique to undertake this project. Specifically, Weka Naïve Bayes classifier is employed. We downloaded 2000 reviews from Google play, used 70% for training the classifying model and 30% for testing the model. The outcome of the manual labelling of the reviews is that 96% of the reviews are categorized as either satisfaction or effectiveness. This suggests that users tend to talk more about how they like or dislike a mobile app or complain about the ineffectiveness of it. Due to this skewed nature of the data, the classifying model testing part of the study yielded the expected outcome. The model precision is high for both Satisfaction and effectiveness quality characteristics of the ISO 9126 quality in use part, since both of them got large training data sets. However, due to the minimal training reviews received by the safety and productivity categories their precision lags. Therefore, the study suggests that users are more concerned about the effectiveness of an app and how satisfied the use of the app and its features makes them.

Preview Text FSKTM 2018 60 IR.pdf Download (1MB) | Preview

Actions (login required)

View Item