Research On Pattern Recognition And Performance Prediction Based On Learner Behavior

The increasing development of online learning platforms provides an open and flexible learning experience for many learners at home and abroad.However,there are usually high dropout rate and poor learning effect in online learning.It droves the emergence of data mining and analysis for learning ——focus on using data mining technology to conduct multi-dimensional and fine-grained analysis of learner behavior data,effectively identifying learning patterns and learning motivations,and further exploring their impact on learning effects,and improve the accuracy of performance prediction.Existing works mainly have the following three problems.Firstly,existing works lack of understanding of the evolutionary pattern of learners' pay-to-receive matching.Second,it lacks of works to identifying learning patterns and predicting learning motivations based on the dynamic evolution of learning efficiency.Finally,it ignores the influence of learning pattern and learning motivation on learning effect,and cannot effectively improve the accuracy of performance prediction.To address the above issues,we proposes the problem of pattern recognition and performance prediction based on learner behaviors.It aims to identify learning pattern and motivation by multi-dimensional and fine-grained analysis of learning behaviors,and construct a deep learning model to predict learning performance more accurately.This paper main research contents and innovations are as follows:(1)We propose a novel problem of pattern recognition and performance prediction based on learner behavior.In order to solve the problem,firstly,we define the calculation formulas for learning effort,learning gains and learning efficiency based on learning behaviors.Then,based on the dynamic evolution of learning efficiency,the Gaussian mixture model is used to recognize learning patterns,and the Capsule network model is used to predict learning motivation.(2)We conduct multi-dimensional and fine-grained learning mining based on eight real course datasets.We analysis four RQ by learning data mining and analysis,in order to explore the relationship between learning pattern,learning motivation and learning behavior pay and gain.(3)We propose a performance prediction model based on learning pattern and learning motivation,referred to as GMM+BLAT.We also evaluate the proposed model on real datasets of eight courses and the results show that they lead to significant improvements compared with the baselines.For example,the improvements in the course of History of Architecture(2)datasets are about 5%-29%,5%-23%,and 3%-7%,5%-29% on the MF,WP,WR,and WF.