Revolutionizing C Programming Language Assessment through The Cognitive Code Profiling Model

Abstract

Mastering the C programming language is a challenge for beginner students, requiring not just syntactical knowledge but deep cognitive engagement. Despite this complexity, traditional assessment methods remain largely superficial, focusing heavily on output correctness while failing to measure the cognitive processes and the specific depth of a student’s understanding. This research addresses this by introducing the Cognitive Code Profiling Model (CCPM), a hybrid assessment framework that synergizes structural code profiling with Natural Language Processing (NLP) to map programming performance to Bloom’s Taxonomy. This approach effectively translates raw code Line-of-Code (LoC) features into measurable indicators of cognitive complexity. The study utilizes a quantitative methodology where static code metric specifically Cyclomatic Complexity are fused with token-based logic (TF-IDF) to create a multi-dimensional feature vector. Validated on a dataset of 348 student submissions, the CCPM achieves a classification accuracy of 96.25% using a Random Forest architecture. Results demonstrate a strong positive correlation between logical complexity features and cognitive depth. This hybrid approach acts as a high-precision diagnostic tool, allowing lecturers to pinpoint specific learning deficits. By shifting the focus from error detection to cognitive profiling, the CCPM enhances the pedagogical strategy for computer science education.