Chatbot Helpdesk for Centralized Meter Record Application (ACMT) by PLN using Deep Learning

The Centralized Meter Record Application (ACMT) is a mobile and back-office application used by field meter recording officers. Users often face issues and wait for responses from helpdesk and technical support. This research aims to develop a chatbot model that can assist helpdesk and technical support in answering questions according to standard operating procedures by using deep learning. A chatbot is a computer program designed to interact with humans through text conversations. Deep learning is chosen for chatbot analysis because it excels in understanding and generating human-like language, enabling natural, context-aware, and adaptable interactions. Data is collected from IT Service Management (ITSM) and used for intent creation. TensorFlow is used for machine learning and deep learning, and the model training process uses itertools library to determine the number of neurons in each layer. The model is trained for 90 epochs, achieving an accuracy of 96% with a model size of 209.3 KB. Testing using the Robot Framework showed an accuracy of 80%.