Adaptive User Interface Design for Volleyball Learning Apps: Empirical Evidence from Google Play Reviews and Mobile Screen Analysis

Adaptive sports-learning apps should not expose the same interface to a first-session learner and to a self-directed advanced player. This paper develops an empirical design basis for a volleyball learning app that adapts information density, feedback granularity, and interaction complexity to learner level. We combined two public datasets: a Google Play review corpus and the MASC mobile UI dataset. From 63,340 reviews we constructed a learning-and-training subset of 7,328 reviews from 1,702 apps and benchmarked five sentiment models using grouped 3-fold cross-validation. The best review classifier was a Linear SVM (accuracy = 0.786, macro-F1 = 0.548). We then compared LDA, NMF, and KMeans topic models on 1,527 negative reviews; LDA with four topics produced the best coherence. The resulting themes were core reliability and login/access friction (34.1%), control quality and monetization friction (28.0%), update/privacy/state-persistence failures (23.1%), and support, billing, and account recovery (14.7%). Novice-coded complaints were 3.74 times more likely than advanced-coded complaints to focus on support and recovery, whereas advanced-coded complaints over-indexed on control and monetization. On MASC, a fusion Linear SVM combining keywords and numeric UI features achieved a macro-F1 of 0.938 for 10-class screen prediction. Complexity clustering produced two stable screen regimes: a low-complexity cluster dominated by Welcome, Login, and Home screens and a high-complexity cluster dominated by List, Menu, and Search screens. Based on these results, the paper specifies a two-mode adaptive volleyball UI blueprint: a guided beginner mode and a dense expert mode. The study shows how reproducible review mining and screen analytics can directly inform adaptive UI decisions for skill-learning applications.