Although millions of mobile apps have been published in the app store, the majority are seldom downloaded or used. This phenomenon has intensified the competition among service providers for user loyalty. There were plenty of studies investigating user loyalty in the mobile-app context; nevertheless, most failed to identify those true loyalty users who embraced attitudinal and behavioral loyalty. To address this research gap, this study aims to understand and predict user loyalty by an extended RFD model. We propose that recency, frequency, and duration are able to reflect behavioral loyalty, while category frequency rate and category duration rate are representations of attitudinal loyalty. Using the actual data collected from a third-party app, we calculate the weights of each variable through the entropy weight method, evaluate users' loyalty in two dimensions, and classify users into four groups (i.e., true loyalty, latent loyalty, moderate loyalty, and no loyalty). We also conduct a dynamic analysis to investigate how users move across different loyalty conditions. The results indicate that majority of users tend to stay on their initial loyalty conditions. For those who have changed their loyalty conditions, it is found that true loyalty users are more likely to become latent loyalty users.

中文翻译：

---

忠诚可以被追求和实现吗？用于理解和预测用户对移动应用程序忠诚度的扩展 RFD 模型

尽管应用程序商店中已经发布了数百万个移动应用程序，但大多数很少被下载或使用。这种现象加剧了服务提供商之间对用户忠诚度的竞争。有大量研究调查了移动应用程序环境中的用户忠诚度；然而，大多数人未能确定那些接受态度和行为忠诚的真正忠诚用户。为了解决这一研究空白，本研究旨在通过扩展的 RFD 模型来理解和预测用户忠诚度。我们建议新近度、频率和持续时间能够反映行为忠诚度，而类别频率率和类别持续时间率是态度忠诚度的表示。使用从第三方应用程序收集的实际数据，我们通过熵权法计算每个变量的权重，评估用户的 忠诚度两个维度，将用户分为四类（即真实忠诚度、潜在忠诚度、中等忠诚度和不忠诚度）。我们还进行了动态分析，以调查用户如何跨越不同的忠诚度条件。结果表明，大多数用户倾向于保持最初的忠诚度条件。对于那些改变了忠诚条件的人，发现真正的忠诚用户更有可能成为潜在的忠诚用户。