An Empirical Study on Managing Energy and Accuracy Requirements of Location Based Android Applications
This research paper was accepted by and presented at the International Conference on Software Engineering & Knowledge Engineering (SEKE) 2019. Our research paper is available here.
The improper use of GPS and location-related APIs may result in abnormal battery drain in Android applications. Over the last few years, the developers’ discussions on improving energy efficiency have increased. In this paper, we mine StackOverflow to analyze and summarize the characteristics of developers’ discussions of managing energy and accuracy-related requirements of location-based Android applications. We extracted 11,911 questions from StackOverflow and filtered 320 relevant questions to answer four research questions. We conducted a manual thematic analysis of relevant questions. Our study shows that the developers are concerned about energy consumption, but are unclear about their preferences as energy and accuracy evolved as conflicting requirements.
The data presented in StackOverflow is available as a data dump at the StackExchange Data Explorer. We extracted the data by composing and running an SQL Query. The query includes the keywords android, location, gps used on the tag field of the dataset. The query produced a csv file Dataset-Query-Results containing 11,911 questions. The dataset contains information about Title, Body, Accepted Answer, Score, Views Count, Favorites Count, Created Date and other relevant information. The second step is a semi-automated method to filter the questions that are specific to energy and accuracy-related issues of location-based Android applications. During this filtering process, we used keyword matching on the Body field of the questions and obtained a filtered dataset with 651 relevant questions. The questions with negative and zero scores were removed from successful and ordinary categories as they were insignificant to our study. This reduced the dataset to 399 relevant questions. We manually read the title and body fields of the questions to verify its relevance to energy and accuracy-related requirements during which 79 false positives were found and removed resulting in 320 questions being considered for thematic analysis. The popularity of a question, used for answering the presented research questions, was calculated using a simple python program.
- Marimuthu C (cs15fv08.muthu@nitk.edu.in)
- Sanjana Palisetti (sanjana.palisetti@gmail.com)
- K. Chandrasekaran (kchnitk@ieee.org)