Penerapan Metode TF-IDF dan Deep Neural Network (DNN) pada Chatbot Layanan Informasi Akademik Fasilkom-TI USU

Abstract

One of the roles of educational institutions is to meet the information needs of students. The availability of valid information that is easily accessible is very vital in the field of education, particularly in the university environment. Dissemination of information in the Fasilkom-TI USU is still carried out in a simple way via WhatsApp messages which are usually shared by komting. This method makes students vulnerable to missing information and makes existing information not properly documented. Apart from that, time and distance limitations are obstacles to asking the campus for information. Thus, there is a requirement for a chatbot as a platform that can provide the information needed and can answer students' questions in an interactive way that can be accessed easily anywhere and at any time. In this research, Term Frequency - Inverse Document Frequency (TF-IDF) and Deep Neural Network (DNN) were used. TF-IDF is used as a word weighter to get meaningful key words from the dataset and then the results are processed using the Deep Neural Network method so that the chatbot can then predict the answer to the user's question. The accuracy produced in this research was 88.67%, indicating that the chatbot system built using TF-IDF and DNN in this research had quite good performance.