XÂY DỰNG VÀ THỰC NGHIỆM PHẦN MỀM HỖ TRỢ ĐÁNH GIÁ CHẤT LƯỢNG ĐỀ THI THEO HƯỚNG ĐỊNH LƯỢNG TẠI TRƯỜNG ĐẠI HỌC CÔNG NGHIỆP QUẢNG NINH

In the context of higher education institutions increasingly reforming assessment practices in alignment with learning outcomes, developing a tool to support the evaluation of examination quality based on scientific and objective criteria has become necessary. This paper presents the development and experimental implementation of software designed to support the evaluation of course examination quality at Quang Ninh University of Industry based on Classical Test Theory (CTT). The software employs difficulty, discrimination, and Cronbach’s Alpha reliability indices; percentile ranking is used to identify high-performing and low-performing groups when calculating the discrimination index. The software was developed in Python, using Pandas and NumPy for data processing and Streamlit for the user interface. The software supports the validation and standardization of input data, calculation of indices, classification of test items, identification of cases requiring review, and generation of analytical reports. Through these functions, the software contributes to improving the effectiveness of assessment activities and provides evidence to support the evaluation of course learning outcomes at Quang Ninh University of Industry. The software was tested on 41 examinations from 29 courses across different academic programs in the first semester of the 2025–2026 academic year, including 4 multiple-choice examinations, 19 essay examinations, and 18 mixed-format examinations. The experimental results show that the software can process all three examination formats, automatically calculate the relevant indices, and provide information for identifying items that are overly easy, overly difficult, have low discrimination, or show abnormal patterns. However, for classes with fewer than 10 students, the discrimination index should be regarded only as a reference indicator. Decisions on whether to retain, revise, or remove an item should be based on a combined consideration of difficulty, discrimination, reliability, and the professional judgment of lecturers and academic departments.