Integrating User Feedback to Enhance Software Quality and User Satisfaction in Mobile Application Development
Bibliographic record
Abstract
Enhancing software quality and user satisfaction in mobile application (app) development is a multifaceted challenge that requires effective user feedback integration. As mobile apps become increasingly central to users' daily lives, ensuring high quality and user satisfaction is vital for apps aiming to maintain a competitive edge. User feedback serves as a direct channel to understand user needs, preferences, and pain points, making it an invaluable resource for continuous improvement and innovation. However, integrating user feedback into the development process remains challenging due to the sheer volume and unstructured nature of the data. The goal of this thesis is to investigate how user feedback from app reviews can be effectively utilized to improve software quality and user satisfaction. We achieve this through three interconnected studies, each targeting a specific sub-goal.\nThe first study aims to examine whether code quality impacts user feedback. Enhancing the code quality of an app normally contributes to an improvement in the app/software quality. We analyzed nine open-source Android apps using a Mining Software Repository approach. We assessed the impact of code quality indicators—such as code smells, readability, and source code complexity—on user feedback, including ratings, sentiment, and toxicity from Google Play Store reviews. Contrary to our expectations, the analysis did not reveal any statistically significant relationship between code quality and user feedback, highlighting the complexity of this relationship and suggesting the need for further investigation.\nThe second study aims to create a refined classification system for user reviews by correlating them to factors such as code quality, software artifacts, and user sentiment. To achieve this, we employed open coding to categorize issues in app reviews into types such as Crashing, Design, Functionality, and Performance. By connecting these categories with code quality metrics and software artifacts, we sought to gain deeper insights into user feedback. Using machine learning models, we automatically classified issues in reviews from seven open-source Android apps, with the fine-tuned Generative Pre-trained Transformer (GPT-3.5) model achieving the highest accuracy at 95%. This study revealed a statistically significant relationship between the classified issues and both code quality metrics and software artifacts, underscoring the complexity of integrating user feedback into the development process and demonstrating the value of automation for managing and prioritizing large volumes of app reviews.\nGiven the statistically significant relationship identified in the second study, we recognized the need to convey these insights to developers through an efficient and practical tool. Therefore, The third study aims to develop a dashboard designed to streamline user feedback management. This dashboard features the GPT-3.5 model trained in the previous study to detect and categorize issues in user reviews, alongside sentiment and toxicity analysis to gauge user emotions and potential toxic feedback. Additionally, it includes code analysis to identify code smells across different app versions. A survey of app developers was conducted to evaluate the dashboard's usability and effectiveness, indicating positive results. 78% of participants reported that the dashboard effectively helped them manage user feedback and monitor code quality, and 84% indicated that they would recommend it to other developers.\nIn this thesis, we explored the impact of code quality on user feedback, developed a classification system for app reviews, and created a practical tool for developers to provide them with comprehensive insights into effectively utilizing user feedback in app development. Our studies highlight the importance of integrating user feedback through automated classification and real-time sentiment and toxicity analysis, ultimately enhancing software quality and user satisfaction.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.018 | 0.087 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".