基于协同过滤的高校个性化就业推荐系统研究

发布时间：2018-01-30 21:10

本文关键词： 个性化就业推荐协同过滤 k-means 信息增益率高校毕业生　出处：《昆明理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着大学生毕业生不断增加,就业难的问题日益凸显。而大学生不能掌握有效就业信息、就业目标定位不准、准备不足都会对就业造成不利影响。面对就业网站大量的招聘信息以及学生和企业间的信息不对称,使得毕业生难以搜索到适合自己的就业单位,只能盲目参加招聘会,这不仅浪费了时间和精力,甚至错过了适合自己的就业机会,大大降低了就业成功率。面对现状,虽然各高校都针对性的开展了就业指导和推荐,但是由于毕业生人数较多,高校无法针对每个毕业生的特点进行个性化推荐,而目前高校的就业网站只能发布就业信息,无法为毕业生推荐适合的就业单位。因此,我们需要寻找一种客观、个性化且能针对个人情况进行推荐的方法和手段。随着个性化推荐系统的研究和应用,为解决毕业生个性化就业推荐问题提供了有利支持。个性化就业推荐系统通过挖掘学生就业意向、职业兴趣、在校表现等多方面信息,结合往届生就业数据,能够为毕业生推荐适合的就业单位,引导毕业生进行有效的就业准备,减少时间和精力的浪费,提高就业成功率。目前在毕业生个性化就业推荐系统方面的研究尚不成熟,推荐效果有待提高,推荐模型和推荐算法仍需改进。本文对目前研究中存在的下述3个不足进行了研究：(1)缺乏结合学生就业特征的推荐。目前使用较多的是传统协同过滤算法,仅仅依靠学生就业兴趣评分,没有考虑到学生特征对于就业的影响。(2)不能客观的确定就业特征的影响权重。目前,特征权重确定大都采用主观评价方法,难以体现客观实际。(3)采用K-means聚类提高推荐速度时没有解决该算法受初始聚类中心影响的问题。针对问题,本文首先分析了影响毕业生就业的因素,从中提取了9个学生就业特征；其次,通过对比分析选择了信息增益率作为计算特征权重的方法；然后,为避免聚类效果受初始聚类中心的影响,提出了改进的AK-means (Adcanced-K-Means)算法对学生特征聚类,利用MATLAB编程验证了算法的有效性,最终,结合学生就业特征和兴趣评分,构造了基于学生特征的协同过滤就业推荐模型。在推荐模型建立的基础上,采集了某工科学院4年的学生就业数据对模型进行验证分析,并采用SQL Sever 2008数据库和C#编程语言开发了基于B/S结构的就业推荐系统原型,该原型能够实现本文模型的推荐功能。通过验证,本文提出的就业推荐模型具有一定的有效性,能够为学生就业提供一定的参考作用,对于推荐系统在高校的应用具有积极的探索意义。
[Abstract]:With the increase of college graduates, the employment problem has become increasingly prominent. The students can't master the effective employment information, employment target positioning, lack of preparation may have an adverse impact on employment. Facing the employment site a lot of recruitment information and the information asymmetry between students and enterprises, makes it difficult for graduates to search for their own employment units, only blind to participate in the recruitment, which is not only a waste of time and energy, even missed their own jobs, and greatly reduce the success rate of employment. The face of the status quo, although all colleges and universities to carry out employment guidance and recommendations, but because the number of graduates in Colleges and universities are not according to the characteristics of each graduate's recommendation, and at present, college employment website can publish employment information, to recommend suitable employment for graduates. Because of this, we need to To find an objective, personalized and ways and means for personal recommendation. With the research and application of personalized recommendation system, and provide favorable support for solving the employment problem. Graduates personalized recommendation personalized employment recommendation system by mining employment intention, student occupation interest, school performance and other aspects of information, combined with the previous employment the data can be recommended for employment of graduates, graduates of effective employment preparation, reduce the waste of time and energy, improve the success rate of employment. At present in the graduates employment recommendation system research on personalized recommendation is not mature, the effect needs to be improved, the recommended model and recommendation algorithms still need to be improved. This paper made a research on the present research in the following 3 aspects: (1) the lack of employment characteristics of students recommended. Currently used more is traditional Collaborative filtering algorithms rely solely on students' Employment Interest score, without considering the characteristics of the students for employment effects. (2) effects of the employment characteristics determine the weight can not be objective. At present, most of the feature weights and subjective evaluation method, is difficult to reflect the objective reality. (3) using the K-means clustering algorithm does not solve the initial clustering center influence of increasing the recommended speed. To solve the problem, this paper first analyzes the factors affecting the employment of graduates, extracted 9 students employment characteristics; secondly, through the comparative analysis of the choice of the information gain rate as a method of feature weight calculation; then, in order to avoid the clustering effect is influenced by the initial cluster center, put forward the improved AK-means algorithm (Adcanced-K-Means) on the characteristics of the student clustering, using MATLAB programming to verify the effectiveness of the algorithm, finally, combined with the characteristics of students' employment and interest Score, structure of the collaborative filtering recommendation model based on the characteristics of students' employment. Based on the model, a collection of Engineering College Students' employment data for 4 years to verify the analysis of the model, and the development of the B/S structure of the employment recommendation system prototype based on using SQL Sever 2008 database and C# programming language, the recommendation function the prototype can realize this model. Through the verification, the proposed employment recommendation model has certain validity, can provide reference for the employment of students, the recommendation system is very significance to explore in the application.

【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：G647.38

【参考文献】