Research On The Intelligent Evaluation Method Of Students’ Ability In Bayesian Network Based On Learning Analysis

In recent years,scholars at home and abroad have paid a lot of attention to the field of learning,and learning outcome prediction is one of the most difficult points.Many scholars have carried out multi-level and multi-dimensional research on it.Traditional teaching aims to cover knowledge points,and most of the existing learning assessment research focuses on the knowledge level and analyzes the mastery of knowledge points.We know that the process of learning is the process by which learners transform knowledge into individual cognition and apply it.Today’s learning analysis research mostly takes knowledge acquisition as the learning goal,and rarely goes deep into the research on the improvement of learners’ cognitive ability and potential trait changes.But in the learning process,the two are closely linked and inseparable.In order to effectively predict students’ learning outcomes,and comprehensively consider the knowledge accumulation and potential trait changes generated by learning,this research has conducted in-depth research on knowledge ability network and learning outcome prediction,including the following two aspects:(1)A model combining knowledge space and ability space is proposed,using knowledge keywords and ability dimensions as the basic elements,to realize the quantitative description of multi-dimensional output of students’ learning in a specific knowledge space;A Bayesian network with "keyword + ability" as the node is constructed to predict students’ learning results through probabilistic deduction.In the research,a priori network of experts is first constructed,and then after data collection and cleaning,the Bayesian network structure is learned through statistics,and the node conditional probability table is obtained,and single student evaluation and prediction are carried out through the discretized mastery level.Further,the dynamic Bayesian network is constructed with the test phase as the time series,and the learning evaluation is carried out from the trend of the overall ability of the class.The experimental results show that the prediction and evaluation effect of the Bayesian network is better than that obtained by the direct evaluation of the traditional item response theory,and the obtained trend of class ability mastery can provide teachers with effective teaching plan adjustment suggestions.(2)A Bayesian knowledge tracking model based on latent features is constructed.In recent years,the research on knowledge tracking model is developing rapidly,but there are still some problems in the knowledge tracking model,among which the lack of data features is a major problem.In order to solve this problem,this study proposes a new latent feature extraction method combined with the real learning process,and adds latent features to the learning process of the knowledge tracking model.Adjust to get the final learning result.The experiment is based on the public knowledge tracking big data set.By performing data analysis and feature selection on the data set,and adding potential feature data,a Bayesian knowledge tracking model including forgetting,slippage and other parameters is constructed.The results show that based on the potential compared with the original Bayesian knowledge tracking model,the feature-based Bayesian knowledge tracking model has better prediction performance,and at some nodes,it also has better prediction performance than some deep knowledge tracking models in recent years.This study provides two models for learning evaluation and learning prediction.The evaluation and prediction results obtained through the model can help teachers in the teaching process,and play an auxiliary role in personalized teaching and the formulation of the overall teaching direction of the class.Some methods and suggestions are provided for ability-oriented intelligent education.