Virtual learning environment to predict withdrawal by leveraging deep learning

Abstract

The current evolution in multidisciplinary learning analytics research poses significant challenges for the exploitation of behavior analysis by fusing data streams toward advanced decision-making. The identification of students that are at risk of withdrawals in higher education is connected to numerous educational policies, to enhance their competencies and skills through timely interventions by academia. Predicting student performance is a vital decision-making problem including data from various environment modules that can be fused into a homogenous vector to ascertain decisionmaking. This research study exploits a temporal sequential classification problem to predict early withdrawal of students, by tapping the power of actionable smart data in the form of students' interactional activities with the online educational system, using the freely available Open University Learning Analytics data set by employing deep long short-term memory (LSTM) model. The deployed LSTM model outperforms baseline logistic regression and artificial neural networks by 10.31% and 6.48% respectively with 97.25% learning accuracy, 92.79% precision, and 85.92% recall. K E Y W O R D S classification, deep learning, long short-term memory (LSTM), students-at-risk. smart data, virtual learning environment (VLE) Int J Intell Syst. 2019;34:1935-1952.wileyonlinelibrary.com/journal/int

Predicting academic performance of students from VLE big data using deep learning models

“…Several studies deploy machine learning techniques to analyse student behavior and predict students at-risk of a failure (Costa et al, 2017;Hassan et al, 2019;Wasif et al, 2019). In the existing literature, another array of studies follow a sequential approach to convert the course duration into a week-wise format and assess student performance according to their interaction with the learning environment.…”

Kajian Penggunaan Data Log Mahasiswa untuk Berbagai Permasalahan Analisis Pembelajaran

“…Referensi [6] juga mengangkat permasalahan tentang drop out atau tidak, berdasarkan relasi durasi pemberian tugas dan aktivitas mingguan mahasiswa. Demikian juga dengan beberapa referensi yang lainnya, yang mengangkat permasalahan yang sama, yaitu mahasiswa akan undur diri (withdrawn) atau tidak, baik pada satu modul tertentu atau keseluruhan modul [7]- [10].…”

“…Selain itu, fitur interaksi juga dapat disajikan dalam bentuk sekuensial, seperti halnya deret waktu, sehingga prediksi kinerja atau kecenderungan mahasiswa melakukan drop out dapat dilakukan secara harian atau mingguan. Fitur sekuensial yang digunakan dalam beberapa literatur sangat bervariasi, yaitu jumlah klik harian dari semua sumber daya VLE [14], jumlah klik pada satu sumber daya VLE tertentu (Forum atau OUcontent atau resource VLE) [8], dan ringkasan dari semua sumber daya VLE (setiap sumber daya VLE memiliki atribut tersendiri) dalam mingguan [7]. Di samping itu, untuk meningkatkan kinerja, fitur sekuensial dengan fitur tabel demografis bisa digabungkan [14].…”

“…Metode pembelajaran mesin yang digunakan untuk melakukan analisis pembelajaran dalam pengolahan OULAD sangat bervariasi, antara lain Decision Tree (DT) [5], [9]- [12], [15], [16]; Multilayer Perceptron (MLP)/Artificial Neural Network (ANN) [17], seperti pada [10], [13]; Support Vector Machine (SVM) [18], seperti pada [5], [9], [10]; Logistic Regression (LR) [15], seperti pada [5], [9]; Long Short-Term Memory (LSTM) [19], seperti pada [7], [14]; Gated Recurrent Unit (GRU) [20]; Random Forest (RF) [5], [9]; Adaptive RF (ARF) [21]; Distributed RF (DRF), Gradient Boosting Machine (GBM), dan Generalized Linear Model (GLM) [22]; Gaussian Mixture Models (GMM) [23], seperti pada [6]; dan Time Series Forest (TSF) [24], seperti pada [8]. Selain itu, ada juga yang mengombinasikan dengan metode Principal Component Analysis (PCA) untuk mereduksi fitur sebelum diproses dengan algoritme pembelajaran mesin [5] dan mengombinasikan LSTM dengan Convolutional Neural Network (CNN) [25], seperti pada [14].…”

Improving M-Learners’ Performance Through Deep Learning Techniques by Leveraging Features Weights

“…For the last two decades, different ML/DL algorithms have been developed, evaluated and their performance explored in online and M-learning settings [5], [6]. It is crucial to decide which type of ML/DL algorithm to choose for modeling the learning behavior of M-learners as proper learning algorithm increases/decreases the response time of the Mlearning system [7], [8]. The right algorithm also affects the overall performance of the M-learning system.…”