当前位置: X-MOL 学术Educ. Inf. Technol. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Predicting the percentage of student placement: A comparative study of machine learning algorithms
Education and Information Technologies ( IF 4.8 ) Pub Date : 2021-07-02 , DOI: 10.1007/s10639-021-10655-4
Erman Çakıt 1 , Metin Dağdeviren 1
Affiliation  

In recent years, there has been an increase in the demand for higher education in Turkey, where the demand, as in most other countries, exceeds what is available. The main purpose of this research is to develop machine learning algorithms for predicting the percentage of student placement based on the data related to the university’s academic reputation, opportunities of the city where the university is located, facilities and cultural opportunities of the university. When the model accuracy was evaluated on the basis of performance metrics, the Extreme Gradient Boosting (XGBoost) algorithm showed greater predictive accuracy than other machine learning approaches. A sensitivity analysis was performed using the extreme gradient boosting machines algorithm to identify the degree to which the input variables contribute to the determination of the output variable. Five input variables, namely the percentage of student placement at year t-1, the university scientific document score, university Phd programme score, university faculty member/student score, and the percentage of student placement at year t-2 were found to be the most effective parameters. Prediction and sensitivity analysis results obtained in this study can be used in many different ways, such as determining the quotas for universities, allocating resources, and making new regulations.



中文翻译:

预测学生安置的百分比:机器学习算法的比较研究

近年来,土耳其对高等教育的需求有所增加,与大多数其他国家一样,需求超过了现有需求。本研究的主要目的是开发机器学习算法,根据与大学的学术声誉、大学所在城市的机会、大学的设施和文化机会相关的数据来预测学生安置的百分比。当基于性能指标评估模型精度时,极限梯度提升 (XGBoost) 算法显示出比其他机器学习方法更高的预测精度。使用极端梯度提升机算法进行敏感性分析,以确定输入变量对输出变量确定的贡献程度。五个输入变量,即当年学生安置的百分比t-1、大学科学文献分数、大学博士课程分数、大学教员/学生分数以及t-2年的学生安置百分比被认为是最有效的参数。本研究中获得的预测和敏感性分析结果可用于许多不同的方式,例如确定大学配额、分配资源和制定新法规。

更新日期:2021-07-02
down
wechat
bug