Sentimen Analisis Berbasis Aspek terhadap Ulasan Kepuasan Pengguna M-Banking Menggunakan Metode Extra-Trees Classifier

Abstract

Mobile banking is also known as a service that is targeted towards the bank‟s customers to promote a more convenient method of transaction, both remote and away from the bank or online. Mobile banking is a form of development within the world of information technology that also plays the role of further evolving the banking sector by utilizing existing internet services. Mobile banking users or bank customers can provide feedback from their experience of the application use. From the reviews, banks can benefit from information and data in improving the application‟s function and quality. Feedback can be in the form of positive, negative, and neutral feedback, where this type of information gathering is a part of this sentiment. The reviews will then first be processed by categorizing specific topics and keywords within the review itself. This is done to ease banks in improving the application‟s ability to pivot and narrow down certain aspects of a review. The data that is used in this research are users of an m-banking application called Livin‟ by Mandiri. With 2000 data gathered through the process of data crawling from the Play Store application. The preprocessing steps of this research include, data cleaning, data normalization, case folding, filtering (stopword removal), stemming, and tokenizing. In the feature extracting step, TF-IDF is utilized to weigh words into vectors, LDA as a tool to determine topic distribution within the data and continued with Extra-Trees Classifier method as identification. Confusion matrix is used as an evaluation metric and the result have shown that there is 76% accuracy with three product aspects that dominates each sentiment.