Development of a Public Complaint Classification Model to Support E-Government Using IndoBERT

The Indonesian government has placed significant focus on the implementation of e-government over the past decade to improve public services through the use of information and communication technology. E-government facilitates public access to government services online, reduces the need for face-to-face interactions, and simplifies the management of various administrative tasks. One of the key aspects of e-government is its ability to respond to public complaints quickly and accurately, thereby improving the quality of public services and increasing public trust in the government. However, approximately 41.4% of the public still feels dissatisfied with the services provided by the government. Therefore, efforts are needed to enhance public trust through digital transformation, with an emphasis on increasing transparency and accountability in service delivery. This research focuses on developing a community complaint classification system utilizing deep learning technology, specifically the IndoBERT model. The dataset used in this study consists of 1,000 augmented public complaint data entries sourced from Satu Data Indonesia. The evaluation results show that the model achieved an accuracy of 99.9%, with a precision of 99.99%, a recall of 99.9%, and an F1-score of 99.89%. The model has proven to be highly effective in helping the government efficiently and accurately categorize and direct public complaints to the relevant agencies. This research, therefore, makes a significant contribution to the development of a more responsive and transparent e-government system for addressing public complaints.