FROM DESCRIPTION TO PREDICTION - UNVEILING STUDENT PERFORMANCE IN ONLINE LEARNING THROUGH DATA-DRIVEN ANALYSIS AND MACHINE LEARNING - Master Thesis in Management

Aug 4, 2026

Upload by Skyler

This thesis investigates how student engagement evolves in an online learning environment during the COVID-19 pandemic. Using a dataset covering student activities, class information, and teacher details from an online platform, the work conducts a staged analytics process. It first performs descriptive analysis of class behavior with clustering, exploratory data analysis, and OLS focused on active and passive learning methods. Next, it applies similar techniques to student-level behavior. Finally, machine learning models (KNN, decision trees, Ridge/Lasso regression, neural networks) power a recommendation system predicting student grades and informing class or teacher assignments, followed by managerial strategies derived from interviews.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages139

Reading Time5 h 50 min

AccessFree to download

Abstract

List of Figures

| FROM DESCRIPTION TO PREDICTION: UNVEILING STUDENT PERFORMANCE IN ONLINE LEARNING THROUGH DATA-DRIVEN ANALYSIS AND MACHINE LEARNING
MASTER THESIS IN MANAGEMENT
ENGINEERING (ANALYTICS FOR
BUSINESS)
Authors: Farbod Forouhideh – Hamed Aliakbarimajid | | | --- | --- | | Student IDs:
Advisor:
Co-advisor:
Academic Year: | 10824619-10818226 Prof. Tommaso Agasisti Prof. Melisa Diaz 2023-24 | | | | ii Abstract Abstract This thesis investigates the dynamics of student engagement in an online learning environment during the COVID-19 pandemic. Leveraging a dataset encompassing student activities, class information, and teacher details on an online platform, the study encompasses a comprehensive exploration. The initial phase involves a literature review encompassing Learning Management Systems (LMS), E-learning, and the impact of COVID-19 on education. The practical component is structured into three segments. The first entails a descriptive analysis of class behavior through Cluster analysis, Exploratory Data Analysis (EDA), and Ordinary Least Squares (OLS) models mainly focusing on active and passive learning methods. Subsequently, a similar approach is applied to analyze student behavior in the second phase focusing on students rather than classes. The final step involves predictive analysis, employing machine learning models such as KNN, Decision Tree, Ridge Lasso Regression, and Neural Networks to create a recommendation system. This system predicts student grades, aiding in the identification of suitable class or teacher assignments. The thesis concludes with practical managerial implications derived from an interview, proposing strategies to enhance both active and passive learning for students in E-learning platforms. Keywords: E-learning, COVID-19, Active and Passive learning, Machine learning, Clustering, Descriptive Analysis, Predictive Analysis. ii iii Abstract in lingua italiana Abstract in lingua italiana Questa tesi indaga le dinamiche del coinvolgimento degli studenti in un ambiente di apprendimento online durante la pandemia di COVID-19. Sfruttando un set di datiche comprende le attività degli studenti, le informazioni sulle lezioni e i dettagli sugli insegnanti su una piattaforma online, lo studio comprende una esplorazionecompleta. La fase iniziale prevede una revisione della letteratura che comprende i Sistemi di Gestione dell'Apprendimento (LMS), l'e-learning e l'impatto del COVID-19 sull'istruzione. Il componente pratico è strutturato in tre segmenti. Il primo comporta un'analisi descrittiva del comportamento delle classi attraverso l'analisi di clustering, l'analisi esplorativa dei dati (EDA) e modelli di regressione lineare ordinaria (OLS) focalizzati principalmente sui metodi di apprendimento attivo e passivo. Successivamente, un approccio simile viene applicato per analizzare il comportamento degli studenti nella seconda fase, concentrando l'attenzione sugli studenti piuttosto che sulle classi. Il passaggio finale comporta un'analisi predittiva, impiegando modelli di machine learning come KNN, alberi decisionali, regressione Ridge Lasso e reti neurali per creare un sistema di raccomandazione. Questo sistema predice i voti degli studenti, aiutando nell'individuazione di assegnazioni di classe oinsegnanti adatti. La tesi si conclude con implicazioni manageriali pratiche derivateda un'intervista, proponendo strategie per migliorare sia l'apprendimento attivo chequello passivo per gli studenti nelle piattaforme di e-learning. Parole chiave: E-learning, COVID-19, Apprendimento attivo e passivo, Machine learning, Clustering, Analisi descrittiva, Analisi predittiva. iii iv List of Figures List of Figures Figure 1) General Overview of Initial Dataset ....................................................................4 Figure 2) General Overview of Initial Dataset ..................................................................21 Figure 3) Frequency of Distribution

FAQs

What problem does the thesis address in online learning during COVID-19?

It examines how student engagement dynamics unfold in an online learning environment during the COVID-19 pandemic.

Which methods are used for the descriptive analysis phase?

Clustering, exploratory data analysis (EDA), and ordinary least squares (OLS) are used to study class behavior, mainly for active and passive learning methods.

How is student performance predicted in the thesis?

A predictive analysis stage uses machine learning models such as KNN, decision trees, Ridge/Lasso regression, and neural networks to build a recommendation system that predicts student grades and supports class or teacher assignment decisions.

Aug 4, 2026

Upload by Skyler