Resumen
Application Programming Interface (API) usage mining is an approach used to extract the common API usage to help developers get used to the APIs. However, in Android applications, the usage can be hidden or fragmented due to class inheritance. Such hidden or fragmented usages could decrease the coverage and accuracy of the existing API mining approaches. Our method aims to resolve the problem of hidden and fragmented usages through API generalization. This generalized usage is expected to be applicable to every class that inherits a class in the usage. In the experiment, among 442,809 Android API usages, 104,839 usages either were hidden or fragmented. By revealing such usages, the accuracy of the code completion was improved by at most 6.66%. The usage generalization was efficient for extracting API usages in Android applications in which the APIs are used through class inheritance.