Discussion-record-based prediction model for creativity education using clustering methods

The development of creativity is a major challenge for engineering education. Creativity not only affects originality and innovation but also increases the utility of products and ideas, thus yielding a competitive advantage in the market. Although studies have demonstrated that creativity can be practically assessed through student innovation or responses to creativity questionnaires, few studies have explored how creativity is developed. Therefore, this study developed a model for predicting creativity based on the discussion records of students during creative activities. The model was designed using the Chinese Knowledge Information Processing tokenization system, a term frequency-inverse document frequency method, and support vector machines. The experimental results revealed a model accuracy of 93 %, thus demonstrating the model's feasibility for predicting creativity. Furthermore, the students' key terms were obtained by using k-means clustering with the proposed model, thus yielding valuable supplementary information for exploring the development of creative ideas. These results revealed that data mining has substantial value in creativity education. Therefore, the proposed predictive model is suitable for predicting creativity in learning activities based on discussion records. It can support teachers' assessment of student performance and provision of timely guidance and feedback to enhance creative thinking and stimulate imagination among engineering students.