推荐系统多样性研究及其在就业推荐中的应用

发布时间：2018-04-01 21:04

本文选题：推荐系统　切入点：就业推荐　出处：《山东师范大学》2017年硕士论文

【摘要】：就业推荐系统对于解决就业问题具有良好的效果,因此受到国内外学者广泛关注,取得了丰富的成果。但是,就业推荐领域中,仍然存在以下不足之处有待进一步完善:第一,推荐结果过于单一,用户视野被局限。第二,热门职位被推荐给过多的求职者,降低了求职成功几率。第三,冷门职位得不到有效推荐,损害了招聘企业的利益。本文以就业推荐多样性优化为主要目标,针对上述问题进行了深入的研究。本文主要创新点及贡献如下:(1)针对推荐系统中的个体多样性问题,提出一种基于聚类的个体多样性优化推荐算法。本文针对传统就业推荐算法缺少对个体多样性的考虑,推荐结果过于单一,提出一种基于聚类的个体多样性优化推荐算法。首先,算法针对系统中项目差异度进行计算,充分考虑了项目属性值间的差异;其次,基于项目差异度采用k-means聚类算法对系统中项目进行聚类;然后,基于现有推荐算法获得预测评分矩阵,设置评分阈值,筛选预测评分大于阈值的项目构建用户候选推荐列表;最后,结合项目聚类信息从用户候选列表中获得一组多样性好的项目推荐给用户。实验结果表明,对于用户个体而言,该算法在保证推荐准确率的同时,能有效提高推荐结果的多样性。通过将算法应用于就业推荐原型系统表明,基于聚类的个体多样性优化推荐算法,可有效提高就业推荐的个体多样性与用户满意度。(2)针对推荐系统中的总体多样性问题,提出一种基于二分图网络的总体多样性优化推荐算法。本文针对传统就业推荐算法缺少对总体多样性的考虑,造成系统“马太效应”日益严重,“长尾”职位数量增多的现象,提出一种基于二分图网络的总体多样性优化推荐算法。首先,算法基于现有推荐算法获得预测评分矩阵,设置评分阈值,筛选预测评分大于阈值的项目构建用户候选推荐列表。其次,基于用户候选推荐列表构建推荐二分图。最后,基于构建的推荐二分图,采用置换增广路中匹配边与非匹配边方法,提高推荐总体多样性。实验结果表明,对于系统整体而言,该算法在保证推荐准确率的同时,能有效提高推荐总体多样性。应用于就业推荐领域的基于二分图网络的总体多样性优化推荐算法,可有效提高就业推荐的总体多样性与用户满意度。(3)基于上述两种多样性优化策略,实现了就业推荐原型系统。
[Abstract]:The employment recommendation system has a good effect on solving the employment problem, so it has received extensive attention from scholars at home and abroad, and has made a lot of achievements. However, in the field of employment recommendation, there are still the following shortcomings to be further improved: first, Recommendation results are too single, user horizons are limited. Second, hot jobs are recommended to too many job seekers, reducing their chances of success. Third, bad jobs are not recommended effectively. This paper focuses on the optimization of the diversity of employment recommendation, and makes a deep research on the above problems. The main innovation and contribution of this paper are as follows: 1) aiming at the individual diversity problem in the recommendation system. An optimal recommendation algorithm for individual diversity based on clustering is proposed in this paper. In view of the lack of consideration of individual diversity in the traditional employment recommendation algorithm, the recommendation result is too single. This paper proposes a clustering based recommendation algorithm for individual diversity optimization. Firstly, the algorithm calculates the item difference degree in the system, and fully considers the difference between item attribute values. K-means clustering algorithm is used to cluster the items in the system based on the item difference degree, and then, based on the existing recommendation algorithm, the prediction scoring matrix is obtained, the scoring threshold is set up, and the user candidate recommendation list is constructed by screening the items whose prediction score is greater than the threshold value. Finally, a group of items with good diversity is obtained from the user candidate list by combining the item clustering information. The experimental results show that the proposed algorithm ensures the accuracy of the recommendation for the user at the same time. It can effectively improve the diversity of recommendation results. By applying the algorithm to the employment recommendation prototype system, it is shown that the clustering based individual diversity optimization recommendation algorithm, It can effectively improve the individual diversity of employment recommendation and user satisfaction. This paper presents an optimal recommendation algorithm for population diversity based on bipartite graph network. In view of the lack of consideration of the overall diversity in the traditional employment recommendation algorithm, the "Matthew effect" of the system is becoming more and more serious, and the number of "long tail" posts is increasing. In this paper, a general diversity optimization recommendation algorithm based on bipartite graph network is proposed. Firstly, based on the existing recommendation algorithms, the prediction score matrix is obtained and the threshold is set. Second, build the recommended dichotomy based on the user candidate recommendation list. Finally, based on the constructed recommendation dichotomy, The method of matching edge and mismatch edge in permutation augmented path is used to improve the diversity of recommendation population. The experimental results show that the proposed algorithm not only guarantees the accuracy of recommendation, but also ensures the accuracy of recommendation for the whole system. It can effectively improve the overall diversity of recommendation. The overall diversity optimization recommendation algorithm based on bipartite graph network is applied in the field of employment recommendation. It can effectively improve the overall diversity of employment recommendation and user satisfaction. 3) based on the above two kinds of diversity optimization strategies, the prototype system of employment recommendation is implemented.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【参考文献】