The wide availability of data on students’ academic pathways has enabled the development of advanced analytical methodologies aimed at understanding and optimizing learning experiences. In this context, Educational Process Mining (EPM) emerges as a discipline that integrates process mining, data mining, and learning analytics to study educational processes, allowing the identification of recurring patterns, deviations from expected behaviors, and the formulation of predictions to support academic decision-making. This study applies EPM to the pathways of students enrolled in the Bachelor's degree program in Computer and Automation Engineering at the Università Politecnica delle Marche, with a particular focus on next activity prediction, that is, forecasting the next exam to be taken. The analysis was conducted on the careers of students who graduated on time, considered as a reference for “optimal” behavior, with the aim of developing a predictive model capable of answering the following question: “Given the current state of a student’s study plan, which exam should be taken to maximize the probability of graduating on time?”. The problem was formalized as a multi-class classification task and addressed using deep learning techniques, specifically through a Deep Graph Convolutional Neural Network (DGCNN). In line with this methodological approach, the study pursues a dual purpose: on one hand, to provide students with data-driven recommendation tools; on the other hand, to support academic institutions in identifying critical issues in the educational offer.
L’ampia disponibilità di dati sui percorsi accademici degli studenti ha reso possibile lo sviluppo di metodologie avanzate di analisi finalizzate a comprendere e ottimizzare le esperienze di apprendimento. In questo contesto si colloca l’Educational Process Mining (EPM), disciplina che integra process mining, data mining e learning analytics per studiare i processi che avvengono in ambito educativo, permettendo di identificare pattern ricorrenti, deviazioni dai comportamenti attesi e di formulare previsioni a supporto delle decisioni accademiche. Il presente studio applica l’EPM ai percorsi degli studenti iscritti al corso di laurea triennale in Ingegneria Informatica e dell’Automazione dell’Università Politecnica delle Marche, concentrandosi in particolare sulla next activity prediction, ovvero la previsione del prossimo esame da sostenere. L’analisi è stata condotta sulle carriere degli studenti laureati in tempo, considerati come riferimento di comportamento “ottimale”, con l’obiettivo di sviluppare un modello predittivo capace di rispondere alla seguente domanda: “Dato lo stato attuale del percorso di studi, quale esame dovrei sostenere per massimizzare la probabilità di conseguire la laurea nei tempi previsti?”. Il problema è stato formalizzato come una classificazione multiclasse e affrontato mediante tecniche di deep learning, in particolare attraverso una Deep Graph Convolutional Neural Network (DGCNN). In linea con questo approccio metodologico, la ricerca si pone una duplice finalità: da un lato, fornire agli studenti strumenti di raccomandazione data-driven; dall’altro, supportare le istituzioni accademiche nell’individuazione di criticità nell’offerta formativa.
Raccomandazione del prossimo esame da sostenere: un approccio basato su Process Mining e Deep Learning
CANICATTÌ, DORIANA
2024/2025
Abstract
The wide availability of data on students’ academic pathways has enabled the development of advanced analytical methodologies aimed at understanding and optimizing learning experiences. In this context, Educational Process Mining (EPM) emerges as a discipline that integrates process mining, data mining, and learning analytics to study educational processes, allowing the identification of recurring patterns, deviations from expected behaviors, and the formulation of predictions to support academic decision-making. This study applies EPM to the pathways of students enrolled in the Bachelor's degree program in Computer and Automation Engineering at the Università Politecnica delle Marche, with a particular focus on next activity prediction, that is, forecasting the next exam to be taken. The analysis was conducted on the careers of students who graduated on time, considered as a reference for “optimal” behavior, with the aim of developing a predictive model capable of answering the following question: “Given the current state of a student’s study plan, which exam should be taken to maximize the probability of graduating on time?”. The problem was formalized as a multi-class classification task and addressed using deep learning techniques, specifically through a Deep Graph Convolutional Neural Network (DGCNN). In line with this methodological approach, the study pursues a dual purpose: on one hand, to provide students with data-driven recommendation tools; on the other hand, to support academic institutions in identifying critical issues in the educational offer.| File | Dimensione | Formato | |
|---|---|---|---|
|
D_Canicatti_tesi_PDFA.pdf
non disponibili
Dimensione
2.11 MB
Formato
Adobe PDF
|
2.11 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/22835