A Study On Learning Behavior Pattern Analysis Based On Artificial Intelligence Classroom Observation

The classroom is the primary site of educational ecology and is the place where teaching and learning behaviors most often occur.In classroom education,learners’ learning behaviors are the externalized result of their learning status and have a direct impact on learning outcomes.Exploring learners’ learning behavior patterns can better understand learners’ knowledge construction process and help teachers gain practical knowledge in implementing learning assessment,providing teaching feedback and taking timely teaching interventions,which is conducive to the improvement of classroom teaching quality and teaching effectiveness.Traditional analysis of classroom learning behavior is mainly based on classroom observation and manual coding,which has shortcomings such as over-reliance on experts,complex coding and inefficient analysis.Artificial intelligence technology provides new opportunities for largescale concomitant acquisition and automated intelligent annotation of classroom teaching behaviors.Identifying and analyzing multimodal classroom learning behaviors through artificial intelligence can help to deeply analyze and understand the development mechanism and influencing factors of classroom teaching,thus providing effective support for educators to carry out effective teaching.However,current classroom learning behavior research focuses on classroom environments in online and smart classrooms,as well as analysis and mining of learning behaviors that focus on teachers’ teaching.At the same time,although AI brings great potential for offline classroom learning behavior recognition,most of the relevant studies focus on AI-supported classroom teaching model construction and behavior recognition technology development,and there is a lack of sufficient attention to learning behavior analysis and mining research for traditional offline conventional classrooms.Therefore,this study utilizes an AI-supported classroom behavior analysis method with a large number of real conventional classroom teaching videos as research samples,focusing on learners’ learning behavior patterns,behavioral habits and behavioral differences in the classroom.In this study,17 mathematics classroom videos from a class in W Middle School in Wuhan,Hubei Province,from March 4,2022 to April 1,2022 were selected as the research sample.First,learners were classified according to their learning motivation and initial ability;then,based on the classroom teaching behavior sequence data encoded by artificial intelligence technology,the lagged sequence analysis method was used to discover the classroom learning behavior patterns and behavioral differences of learners with different characteristics;finally,the impact of different classroom learning behaviors on learning effects was explored.The results of the learning behavior pattern analysis showed that learners can be classified into "high motivation and high ability","medium motivation and medium ability" and "low motivation and low ability" according to their motivation and initial ability.The three categories are "high motivation and high ability","medium motivation and medium ability" and "low motivation and low ability".In the teaching activities of new lectures,the classroom learning behavior sequences of learners are more concentrated and have a strong structure,and the classroom learning behavior patterns of learners in different clusters do not differ much from each other.In the lecture-assessment activity,the classroom learning behavior sequences of learners were more dispersed,and the classroom learning behavior patterns differed more among different clusters of learners.High-motivation,high-ability learners had more effective learning behaviors of raising their hands and responding for longer periods of time and stronger connections between effective learning behaviors than did medium-motivation,medium-ability and low-motivation,low-ability learners.Compared to high motivation and high ability learners,medium motivation and medium ability learners also had more hand raising and responding effective learning behaviors,but were less focused and engaged in persistence compared to high motivation and high ability learners.Compared to the high-motivation high-ability and medium-motivation mediumability learners,the low-motivation low-ability learners had less effective learning behaviors of raising their hands and responding,especially in the lecture-assessment sessions,and were less engaged and motivated.The results of correlation and regression analyses showed that hand-raising behaviors were significantly associated with learning outcomes in all clusters,and that teacher-board-student reading and writing behaviors of high motivation and high ability,teacher-student interactionstudent hand-raising behaviors of medium motivation and medium ability,and teacher-rounding-student hand-raising behaviors of low motivation and low ability possessed significant predictive power for learning outcomes.Finally,based on the results of the analysis,this study conducts a deeper investigation with the real environment and conditions,proposes reasonable classroom teaching methods and teaching strategies,provides effective suggestions to help teachers improve classroom learning and learning effectiveness,as well as presents outlooks to provide useful experiences for the ground implementation of AI-supported classroom observation analysis.