DEVELOPMENT OF LEADS QUALIFICATION MODEL FOR THE CHATBOT OF SWISS GERMAN UNIVERSITY

Abstract

This study investigated the potential of deep learning models to automate lead qualification in the context of university admissions. The focus was on Swiss German University's (SGU) current manual system for processing WhatsApp conversations with prospective students. While manual review ensures quality control, it suffers from limitations such as time constraints, subjectivity leading to inconsistencies and the risk of human error. This research explored developing and evaluating deep learning models for lead qualification to address these limitations. Convolutional Neural Networks (CNNs) and a hybrid CNN-LSTM architecture were investigated for their effectiveness in classifying leads based on their level of interest and stage in the enrollment process. The models were trained and evaluated on a dataset of WhatsApp conversations in Bahasa Indonesia. The result of this study is that a CNN-LSTM model achieved a promising accuracy of 80.85%. This suggests significant potential for automation and improved efficiency compared to the current manual approach. Furthermore, by automating lead classification, the model minimizes human error and ensures consistent lead categorization, aiming to streamline the university's student recruitment process.