{"title":"教育中的大数据:学生风险案例研究","authors":"Ahmed B. Altamimi","doi":"10.48084/etasr.6190","DOIUrl":null,"url":null,"abstract":"This paper analyzes various machine learning algorithms to predict student failure in a specific educational dataset and a specific environment. The paper handles the prediction of student failure given the students' grades, course difficulty level, and GPA, differing from most of the provided studies in the literature, where focus is given to the surrounding environment. The main aim is to early detect students at risk of academic underperformance and implement specific interventions to enhance their academic outcomes. A diverse set of eleven Machine Learning (ML) algorithms was used to analyze the dataset. The data went through preprocessing, and features were engineered to effectively capture essential information that may impact students' academic performance. A meticulous process for model selection and evaluation was utilized to compare the algorithms' performance with regard to metrics such as accuracy, precision, recall, F-score, specificity, and balanced accuracy. Our results demonstrate significant variability in the performance of the different algorithms, with Artificial Neural Networks (ANNs) and Convolutional Neural Networks (CNNs) showing the highest overall performance, followed closely by Gradient Boosting Classifier (GBC), Neuro-Fuzzy, and Random Forest (RF). The other algorithms exhibit varying performance levels, with the Recurrent Neural Networks (RNNs) showing the weakest results in recall and F-score. Educational institutions can use the insight gained from this study to make data-driven decisions and design targeted interventions to help students at risk succeed academically. Furthermore, the methodology presented in this paper can be generalized and applied to other educational datasets for similar predictive purposes.","PeriodicalId":11826,"journal":{"name":"Engineering, Technology & Applied Science Research","volume":"55 1","pages":"0"},"PeriodicalIF":1.5000,"publicationDate":"2023-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Big Data in Education: Students at Risk as a Case Study\",\"authors\":\"Ahmed B. Altamimi\",\"doi\":\"10.48084/etasr.6190\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"This paper analyzes various machine learning algorithms to predict student failure in a specific educational dataset and a specific environment. The paper handles the prediction of student failure given the students' grades, course difficulty level, and GPA, differing from most of the provided studies in the literature, where focus is given to the surrounding environment. The main aim is to early detect students at risk of academic underperformance and implement specific interventions to enhance their academic outcomes. A diverse set of eleven Machine Learning (ML) algorithms was used to analyze the dataset. The data went through preprocessing, and features were engineered to effectively capture essential information that may impact students' academic performance. A meticulous process for model selection and evaluation was utilized to compare the algorithms' performance with regard to metrics such as accuracy, precision, recall, F-score, specificity, and balanced accuracy. Our results demonstrate significant variability in the performance of the different algorithms, with Artificial Neural Networks (ANNs) and Convolutional Neural Networks (CNNs) showing the highest overall performance, followed closely by Gradient Boosting Classifier (GBC), Neuro-Fuzzy, and Random Forest (RF). The other algorithms exhibit varying performance levels, with the Recurrent Neural Networks (RNNs) showing the weakest results in recall and F-score. Educational institutions can use the insight gained from this study to make data-driven decisions and design targeted interventions to help students at risk succeed academically. Furthermore, the methodology presented in this paper can be generalized and applied to other educational datasets for similar predictive purposes.\",\"PeriodicalId\":11826,\"journal\":{\"name\":\"Engineering, Technology & Applied Science Research\",\"volume\":\"55 1\",\"pages\":\"0\"},\"PeriodicalIF\":1.5000,\"publicationDate\":\"2023-10-13\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Engineering, Technology & Applied Science Research\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.48084/etasr.6190\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"0\",\"JCRName\":\"ENGINEERING, MULTIDISCIPLINARY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Engineering, Technology & Applied Science Research","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.48084/etasr.6190","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"0","JCRName":"ENGINEERING, MULTIDISCIPLINARY","Score":null,"Total":0}
Big Data in Education: Students at Risk as a Case Study
This paper analyzes various machine learning algorithms to predict student failure in a specific educational dataset and a specific environment. The paper handles the prediction of student failure given the students' grades, course difficulty level, and GPA, differing from most of the provided studies in the literature, where focus is given to the surrounding environment. The main aim is to early detect students at risk of academic underperformance and implement specific interventions to enhance their academic outcomes. A diverse set of eleven Machine Learning (ML) algorithms was used to analyze the dataset. The data went through preprocessing, and features were engineered to effectively capture essential information that may impact students' academic performance. A meticulous process for model selection and evaluation was utilized to compare the algorithms' performance with regard to metrics such as accuracy, precision, recall, F-score, specificity, and balanced accuracy. Our results demonstrate significant variability in the performance of the different algorithms, with Artificial Neural Networks (ANNs) and Convolutional Neural Networks (CNNs) showing the highest overall performance, followed closely by Gradient Boosting Classifier (GBC), Neuro-Fuzzy, and Random Forest (RF). The other algorithms exhibit varying performance levels, with the Recurrent Neural Networks (RNNs) showing the weakest results in recall and F-score. Educational institutions can use the insight gained from this study to make data-driven decisions and design targeted interventions to help students at risk succeed academically. Furthermore, the methodology presented in this paper can be generalized and applied to other educational datasets for similar predictive purposes.