IMPLEMENTASI METODE TF-IDF DAN COSINE SIMILARITY PADA FAQ RETRIEVAL CHATBOT PENERIMAAN MAHASISWA BARU DI STMIK AMIKOM SURAKARTA

This study aims to design and implement a New Student Admission (PMB) FAQ Retrieval Chatbot on the STMIK AMIKOM Surakarta website using Term Frequency-Inverse Document Frequency (TF-IDF) and Cosine similarity. The system was developed to support automated PMB information services by retrieving answers from a validated FAQ knowledge base. The final knowledge base consists of 1,010 FAQs in six categories, producing 5,587 corpus documents and 12,785 TF-IDF vocabulary terms. The system was implemented using WordPress as the user interface, FastAPI as the REST API backend, MySQL as the database, and a Virtual Private Server (VPS) as the backend deployment environment. The chatbot applies a staged Processing mechanism consisting of Rule-based Processing, TF-IDF and Cosine similarity, Fuzzy Matching, and Escalation. Text PreProcessing includes case folding, removal of non-alphanumeric characters, informal-word normalization, and whitespace normalization. An administrative dashboard is also provided for FAQ management, reIndexing, conversation history, lead management, monitoring, and analytics. The regression test involving 30 critical test cases resulted in 30 content PASS results with all routing Layers functioning as expected. The Confusion Matrix evaluation on 50 questions achieved an Accuracy of 82.00%, Precision of 76.92%, Recall of 100.00%, and F1-Score of 86.96%. The 5-fold Cross Validation on 4,170 question variations produced an Exact top-1 Accuracy of 33.02% ± 1.02%, while the Answer-equivalent Top-5 analysis achieved 86.59% ± 1.02%. The User Acceptance Test resulted in 100% functional scenario acceptance, while the USE Questionnaire achieved 87.50%, categorized as Very Feasible. The results indicate that the system can support PMB information services, although its generalization capability for unseen question variations and the management of semantically overlapping FAQs still require further improvement.

Keyword:TF-IDF, Cosine similarity, FAQ Retrieval, Chatbot, New Student Admission.