Classification of Customer Complaints: TF-IDF Approaches

Bank Rakyat Indonesia (BRI) is an Indonesian state-owned bank; the operations of the bank certainly aim to provide satisfaction to customers. This can be realized by quickly responding to the bank's response to complaints. The number of complaints received by BRI banks to date is large, but the categories of complaints are still mixed so each division of the bank has difficulties and must first filter complaints. This study aims to classify bank complaints using the Text Mining field to make it easier for BRI to provide complaints directly to each division. The classification method used is Neural Network and the data preprocessing method uses the TF -IDF method. The dataset used in the form of text is one million datasets with 5 classes, namely debit cards, credit cards, customer service, mobile banking, and loans. The results of this study indicate that the accuracy of the neural network method is better and becomes a recommendation to be implemented at BRI. This research is useful for banking in Indonesia or the service industry in managing customer relationship management.