融合多知识点与群体特征的个性化知识推荐方法  

A personalized knowledge recommendation method integrating multi-knowledge point and group characteristics

作　　者：蔡林沁[1] 刘昱辰 任波 蔡志伟 CAI Linqin;LIU Yuchen;REN Bo;CAI Zhiwei(School of Automation/School of Industrial Internet,Chongqing University of Posts and Telecommunications,Chongqing 400065,P.R.China)

机构地区：[1]重庆邮电大学自动化学院/工业互联网学院,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2024年第5期1023-1031,共9页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家自然科学基金项目(62277008);重庆邮电大学教育信息化重点项目(xxhyf2022-08)。

摘　　要：个性化知识点推荐是智慧教育中的重要问题,旨在提供个性化学习服务和提升学习效果。现有基于深度知识追踪的推荐方法,难以直接处理包含多知识点的综合性习题。仅依赖学习者自身的知识状态进行推荐,存在数据稀疏和结果单一的问题。提出一种多知识融合深度知识追踪与群体特征协同过滤性的个性化知识点推荐方法,将习题对应的多个知识点视为习题特征,引入多热编码表示习题与知识点间的关系,对学习者知识状态建模;通过考虑目标学习者的知识水平和学习者群体的近邻信息,获得相似学习者,并利用协同过滤算法获取群体学习偏好进行知识点推荐。该方法可以更好地适应多知识点习题推荐,同时挖掘群体学习特征,提升推荐效果。Personalized knowledge recommendation is a crucial issue in smart education,aiming to provide personalized learning services and improve learning outcomes.Existing recommendation methods based on deep knowledge tracing have difficulty directly handling comprehensive exercises that involve multiple knowledge points.Relying solely on the learner’s own knowledge state for recommendations poses challenges such as data sparsity and single-result outcomes.This paper proposes a personalized knowledge point recommendation method that integrates multi-points fusion deep knowledge tracking and group characteristics collaborative filtering(MDKT-GCCF).The proposed MDKT-GCCF method treats multiple knowledge points corresponding to exercises as exercise features and introduces multi-hot encoding to represent the relationship between exercises and knowledge points,thus modeling the learner’s knowledge state.Based on the knowledge level of target learner and the neighbor information of the learner group,similar learners are identified,and a collaborative filtering algorithm is used to obtain group learning preferences for knowledge point recommendation.The proposed MDKT-GCCF method better adapts to recommending exercises with multiple knowledge points while also uncovering group learning characteristics,enhancing recommendation effectiveness.

关 键 词：多知识融合 知识追踪 协同过滤 群体特征 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]