基于知识图谱与协同过滤的员工培训推荐研究
余江龙 , 苏治文 , 汪德超 , 李海龙 , 马正忠
云南民族大学学报(自然科学版) ›› 2025, Vol. 34 ›› Issue (05) : 590 -596.
基于知识图谱与协同过滤的员工培训推荐研究
Research on employee training recommendation based on knowledge mapping and collaborative filtering
针对电力系统职工培训内容泛化导致个性化推荐不足的问题,提出了一种基于知识图谱和协同过滤相结合的员工培训推荐算法.首先,根据培训岗位中的能力要求,将员工的知识点划分为6大类知识模块,计算员工的知识模块综合评分,构建员工 - 知识模块评分矩阵.同时,根据员工的职称对其进行分类,作为个性化推荐的重要标签.其次,利用培训评价一体化标准构建岗位知识图谱,并结合员工的职称和评分生成员工知识掌握图谱.最后,结合员工-知识模块评分矩阵、职称标签信息和知识掌握图谱,构建知识图谱和协同过滤相结合的员工培训推荐模型,对员工的培训内容进行个性化推荐,并利用实际培训数据进行验证.结果表明,与LDA主题模型相比,协同过滤推荐算法在各个职称人群推荐精确率、召回率和覆盖率均为最高,整体推荐精确率、召回率和覆盖率分别提升19.29、23.21和5.00个百分点,达到92.86%、94.48%和20.36%;所构建模型可有效解决数据稀疏性问题,实现对员工培训内容的个性化推荐.
Aiming at the problem of insufficient personalized recommendation due to the generalization of staff training content in power system, a staff training recommendation algorithm based on the combination of knowledge map and collaborative filtering is proposed. First of all, according to the ability requirements of training posts, the knowledge points of employees are divided into six categories of knowledge modules, the comprehensive score of knowledge modules of employees is calculated, and the employee knowledge module scoring matrix is constructed. At the same time, employees are classified according to their professional titles as important labels for personalized recommendation.Secondly, the integrated training evalution standard is used to build a job knowledge graph, and employees' professional titles and scores are combined to generate an employee knowledge mastery graph. Finally, combined with the scoring matrix of employee knowledge module, title tag information and knowledge mastery map, an employee training recommendation model combining knowledge map and collaborative filtering was constructed, personalized recommendations were made for employee training content, and actual training data was used for verification. The results show that compared with LDA topic model, collaborative filtering recommendation algorithm has the highest recommendation accuracy, recall rate and coverage rate in each professional title group. The overall recommendation accuracy, recall rate and coverage rate are increased by 19.29, 23.21 and 5.00 percentage points, respectively, to 92.86%, 94.48% and 20.36%; The model can effectively solve the problem of data sparsity and realize the personalized recommendation of employee training content.
协同过滤 / 知识图谱 / 职称标签 / 员工培训 / 个性化推荐
collaborative filtering / knowledge mapping / title label / staff training / personalized recommendation
| [1] |
孙传明,周炎,涂燕.基于混合协同过滤的个性化推荐方法研究[J].华中师范大学学报(自然科学版),2020,54(6):956 - 962. |
| [2] |
|
| [3] |
|
| [4] |
陈婷,朱青,周梦溪, |
| [5] |
|
| [6] |
|
| [7] |
|
| [8] |
|
| [9] |
|
| [10] |
|
| [11] |
|
| [12] |
罗承天,叶霞.基于知识图谱的推荐算法研究综述[J].计算机工程与应用,2023,59(1):49 - 60. |
| [13] |
|
| [14] |
赵晔辉,柳林,王海龙, |
| [15] |
徐孟奇,熊熙,李斌勇, |
| [16] |
牛雪磊,杨军,闫浩文.融合知识图谱与协同过滤的微地图推荐[J].地球信息科学学报,2024,26(4):967 - 977. |
| [17] |
金宇,陈红梅,罗川.基于知识图谱的兴趣捕捉推荐算法[J].计算机科学,2024,51(1):133 - 142. |
| [18] |
张屹晗,王巍,刘华真, |
| [19] |
焦富森,李树青.基于物品质量和用户评分修正的协同过滤推荐算法[J].数据分析与知识发现,2019,3(8):62 - 67. |
| [20] |
李艳燕,张香玲,李新, |
中国南方电网有限责任公司科技项目(YNKJXM20230135)
/
| 〈 |
|
〉 |