Klasifikasi Perubahan Perangkat Lunak pada Mobile App Review dengan Menggunakan Metode Long Short Term Memory (LSTM)

Abstract

Along with the development of mobile applications, software evolution is an essential step to
be done. Mining user experience about the product is one of strategies to obtain many information
about the features. Software change categories that are often used are Bug Error, Feature
Request, and Non Informative. Previous research categorized software changes by analyzing the
similarity of hidden topics produced by Latent Dirichlet Allocation (LDA). But the performance
of labeling is not good enough because it only considers the similarity value of some terms that
represent the review sentence. Therefore, this study proposes a method that considers similarity
clustering value and lexical analysis of the document, and classify using Long Short Term
Memory (LSTM) then. The experimental result shows the best classification for Bug Report
software change categories by reaching 93.1% for accuracy, 100% for precision, and 93.1% for
recall.