Evaluation and Verification of Design Problem-solving Ability Based on Multi-modal Data
Design problem-solving ability is a necessary skill for learners in the 21st century.Accurate assess-ment of it will help cultivate innovative talents for the intelligent age.The research aimed at the multimodal eva-luation of design problem-solving ability,and carried out empirical research on the basis of the existing evaluation framework to verify its effectiveness.The experiment recruited 181 subjects,and obtained the physiological data,interactive text data,and psychological questionnaire data of the subjects in the process of solving design prob-lems.Data mining methods such as BERT algorithm,random forest algorithm and keyword frequency analysis were used to quantitatively characterize the dimensions of reflection adjustment,empathy,viewpoint construction,and organizational coordination.Finally,multiple regression methods were used for data fusion,and problem-solv-ing ability was assessed in a procedural,precise and individualized manner.The results showed that the accuracy rate of 72.5%can be achieved by using multimodal data representation and quantitative design problem solving ability.The research provided new ideas and methods for multimodal data representation of learners'high-level abilities,and provided strong support for the subsequent realization of automatic evaluation of design prob-lem-solving abilities,and laid the foundation for the future training and intervention of design problem-solving abilities in real teaching environments.
NETL
NSTL
万方数据
