In Italy, Early School Leaving continues to be one of the main challenges of the education system, with lasting effects on the individual well-being and professional opportunities of students. Despite efforts to counter this phenomenon, statistics suggest that many students still fail to complete their educational journey. However, the increasing availability of educational data, resulting from the digitisation of school systems, offers new opportunities to improve student management and intervene early to reduce this risk. This thesis proposes a standardised and replicable protocol for the development of an Early Warning System (EWS) to identify at-risk students in a timely manner, facilitating targeted and proactive interventions. The protocol, the heart of our work, guides every step of the process: from data collection and preparation to the selection and evaluation of predictive models. Through a structured approach, the protocol integrates advanced analytical techniques and machine learning to predict both students' final academic status and their grades in individual subjects. The thesis explores the Italian educational context, analysing the available data, such as those coming from electronic registers and INVALSI tests, and assessing the applicability of Learning Analytics methodologies in this context. Furthermore, a critical literature review is provided to identify already validated models and approaches that can be adapted and implemented in the national context. The work is part of the DATA2LEARN@EDU project, in collaboration with Politecnico di Milano, and aims to create a replicable protocol for the design of EWS in different educational contexts, helping to improve school outcomes and reduce the phenomenon of early drop-out.
In Italia, l'abbandono scolastico precoce continua a rappresentare una delle principali sfide del sistema educativo, con effetti duraturi sul benessere individuale e sulle opportunità professionali degli studenti. Nonostante gli sforzi per contrastare questo fenomeno, le statistiche suggeriscono che molti giovani non completano il loro percorso di studi. La crescente disponibilità di dati educativi, derivante dalla digitalizzazione dei sistemi scolastici, offre però nuove opportunità per migliorare la gestione degli studenti e intervenire tempestivamente per ridurre questo rischio. Questa tesi propone un protocollo standardizzato e replicabile per lo sviluppo di un sistema di allarme precoce (Early Warning System, EWS), che consenta di identificare in modo tempestivo gli studenti a rischio, facilitando interventi mirati e proattivi. Il protocollo, cuore del nostro lavoro, guida ogni fase del processo: dalla raccolta e preparazione dei dati, fino alla selezione e valutazione dei modelli predittivi. Attraverso un approccio strutturato, il protocollo integra tecniche avanzate di analisi e machine learning per prevedere sia lo stato accademico finale degli studenti sia i loro voti nelle singole materie. La tesi esplora il contesto educativo italiano, analizzando i dati disponibili, come quelli provenienti dai registri elettronici e dalle prove INVALSI, e valutando l’applicabilità delle metodologie di Learning Analytics in tale ambito. Inoltre, viene fornita una revisione critica della letteratura per identificare modelli e approcci già validati che possano essere adattati e implementati nel contesto nazionale. Il lavoro si colloca nel progetto DATA2LEARN@EDU, in collaborazione con il politecnico di Milano, e ambisce a creare un protocollo replicabile per la progettazione di EWS in diversi contesti educativi, contribuendo a migliorare gli esiti scolastici e a ridurre il fenomeno dell'abbandono precoce.
Building an early warning system: a protocol for data collection, analysis, and predictive modelling in education
Boles, Giorgio;Carli, Giorgia
2023/2024
Abstract
In Italy, Early School Leaving continues to be one of the main challenges of the education system, with lasting effects on the individual well-being and professional opportunities of students. Despite efforts to counter this phenomenon, statistics suggest that many students still fail to complete their educational journey. However, the increasing availability of educational data, resulting from the digitisation of school systems, offers new opportunities to improve student management and intervene early to reduce this risk. This thesis proposes a standardised and replicable protocol for the development of an Early Warning System (EWS) to identify at-risk students in a timely manner, facilitating targeted and proactive interventions. The protocol, the heart of our work, guides every step of the process: from data collection and preparation to the selection and evaluation of predictive models. Through a structured approach, the protocol integrates advanced analytical techniques and machine learning to predict both students' final academic status and their grades in individual subjects. The thesis explores the Italian educational context, analysing the available data, such as those coming from electronic registers and INVALSI tests, and assessing the applicability of Learning Analytics methodologies in this context. Furthermore, a critical literature review is provided to identify already validated models and approaches that can be adapted and implemented in the national context. The work is part of the DATA2LEARN@EDU project, in collaboration with Politecnico di Milano, and aims to create a replicable protocol for the design of EWS in different educational contexts, helping to improve school outcomes and reduce the phenomenon of early drop-out.File | Dimensione | Formato | |
---|---|---|---|
2024_12_Boles_Carli_Executive Summary.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Executive summary Boles Carli
Dimensione
925.32 kB
Formato
Adobe PDF
|
925.32 kB | Adobe PDF | Visualizza/Apri |
2024_12_Boles_Carli_Tesi.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Elaborato tesi Boles Carli
Dimensione
5.35 MB
Formato
Adobe PDF
|
5.35 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/230432