Energy inefficiency is a major issue for the application users and they express their concerns using user feedback or app-reviews. That is why developers consistently analyze user feedback to identify recently emerged issues.

We have submitted our work in ACM MobileSoft 2020 where the findings of elaborate empirical study on supervised text classification and topic modeling approaches were reported for isolating energy related reviews from the hoard of feedback and for identifying recently emerged issues. We empirically studied and compared the accuracy, F1-score and run time of 60 machine-learning models with relevant feature combinations and 30 Neural Network-based models developed using six neural network architectures and three word embedding. Furthermore, we experimented the most frequently used string matching with results obtained from applying two of the state-of-the-art topic modeling algorithms in order to compare the techniques and to help the developers investigate the emerging issues responsible for energy inefficiency of the apps.

The proposed interactive visualization tool make it easy for the developers to traverse through the extensive result set generated by the text classification and topic modeling approaches and help them to better comprehend the outcomes of implemented model feature combinations, algorithms used. The data structure used for the tool stores the outcomes of discussed approaches and cumulatively updates itself whenever new developer performs any of the given approaches with a new data-set. This tool also gives developers choices to illustrate the result set as a whole, as a chunk, or as a gradual step-by-step representation.