Automated System Recommends Programming Content via Pattern-based KCs

In introductory programming education, hands-on practice and concise learning activities are crucial for mastering fundamental concepts. While numerous such resources exist, organizing and connecting them in a pedagogically meaningful way typically requires extensive expert curation. This research explores an automated solution using "pattern-based Knowledge Components" (KCs) to identify conceptually similar code-based learning materials. The proposed method involves extracting these pattern-based KCs from individual code samples and then determining the similarity between sets of KCs associated with different activities. This approach leverages alignment at the level of semantically important programming patterns, enabling contextually appropriate and instructionally useful recommendations. The method was evaluated using an expert-curated corpus of introductory Python materials, where instructors had grouped items by conceptual similarity. The results demonstrated that the pattern-based KC approach effectively retrieved resources aligned with this expert organization, outperforming both traditional KC- and embedding-based baseline methods in standard ranking evaluations. This framework promises to provide targeted, concept-oriented guidance for programming learners and assist instructors in efficiently organizing and recommending instructional content.