Pengembangan Prototipe Aplikasi Pembelajaran English Proficiency Test Berbasis Generative AI dengan Pendekatan Adaptif Berbasis Kerangka CEFR

TOEFL and IELTS are among the most widely used assessments for evaluating English language proficiency. However, test-takers often face difficulties in the Structure and Written Expression section of TOEFL and the Writing section of IELTS, which frequently hinder them from reaching their target scores. Traditional learning methods, which lack personalization, further limit the provision of materials tailored to individual needs. To address these challenges, this study proposes the integration of intelligent and adaptive technologies into language learning. Machine learning provides a promising approach for classifying English proficiency levels in accordance with the Common European Framework of Reference for Languages (CEFR). Additionally, Generative Artificial Intelligence (AI) powered by Large Language Models (LLMs) enables the generation of personalized learning content suited to learners’ specific requirements. This study introduces the development of an adaptive learning application for TOEFL and IELTS preparation, built on Generative AI and guided by the CEFR framework. The application was developed using a prototyping approach with iterative refinement to ensure relevance to user needs. Logistic Regression was identified as the most effective model for CEFR-level prediction. Furthermore, usability testing with the System Usability Scale (SUS) yielded a score of 77.5, categorized as “good,” demonstrating the feasibility of the proposed solution.. Keywords—English Proficiency Test, CEFR, Generative AI, Natural Language Processing, Large Language Model (LLM