A novel possibilistic artificial immune-based classifier for course learning outcome enhancement

How Does Learning Analytics Contribute to Prevent Students’ Dropout in Higher Education: A Systematic Literature Review

“…The articles collected referred studies conducted in different countries and continents: Asia (sixteen studies with data retrieved from universities in ten different countries), Europe (thirteen studies with data from universities in six different countries), North America (ten studies with data obtained from universities in two different countries), South America (seven studies concerning data of universities from four different countries), Africa (the study in [77] considers 496 records of data obtained from an online course of Management Informatics on the Abdelmalek Essaadi University, in Morocco), and Australia (the study published in [78] was performed using data about online learning activities of an Australian university). Furthermore, the study described in [87] does not state the university where it is performed, only revealing that it focuses on 16,000 grade records obtained from students. Finally, the study referred to in [88] includes engineering courses of six different universities in five different countries of the European Union.…”

“…We also found that some of the analysed studies had very specific objectives. These include implementing retention strategies [75] (retention here means to convince students to not desert or drop out); identifying student satisfaction [114]; recommending strategies to reduce attrition by reducing dropout [76]; analysing students' learning behaviour through the creation of a feature matrix for keeping information related to the local correlation of learning behaviour [72]; analysing activity, polarity, and emotions of students and tutors to perform sentiment analysis to help in dropout prediction [73]; monitoring the learning process and performing student profiling to support pedagogical actions to reduce dropout [80]; predicting remedial actions [87]; using data to improve courses [96] and learning experiences [96]; and exploring relationships between programming behaviour, student participation, and the outcomes obtained [112].…”

“…However, there are also approaches where the target variable is the students' satisfaction [114], ranking [107], or graduation rates [96]. Some also tried to predict remedial actions to take if the student is at risk of failure [87], or even the urgency in replying to students' posts on the learning management system (LMS) forum [84]. To perform the predictions, the authors took advantage of a set of information (features) that will be described next.…”

“…The process data concern the data obtained during the process of frequenting the program and includes students' achievements and performance [2,71,75,82,87,91,116], interaction with the university's LMS [71][72][73][74][75][76][77][83][84][85][86]93,107,108,113], lecture attendance [90,91,93,94,113] and participation [113], degree [82,85,[87][88][89]92,105,106,109,115,118] and subjects [81,82,85,88,92,106,118] the student is enrolled in, study progression data [104], discipline precedence [117], lecture programme…”

“…There are two algorithms that were used twice in the different studies analysed: convolutional neural networks (CNNs) [72,101], and MLP [76,88]. The following algorithms were used only once in the analysed studies: boosted decision trees [2], Bayes net [76], radial basis function (RBF) [79], the ensembles AdaBoost and LogiBoost [79], PAIRS3 [87], multi-view multiinstance (MVMI) and multi-view Gaussian process (MVGP) [113], and ensemble of linear discriminant analysis (E-LDA) [114].…”