Croatian Journal of Education - Hrvatski časopis za odgoj i obrazovanje, Vol 16, No 2 (2014)

Font Size:  Small  Medium  Large

Using Data Mining to Predict Success in Studying / Primjena rudarenja podataka u predviđanju uspješnosti studiranja

Vlado Simeunovic, Ljubiša Preradović

Abstract


Abstract
This paper deals with the creation of a model for predicting the performance of students during their studies using data mining, as well as with the analysis of factors which affect the achieved level of success. The model that is created on the basis of students’ socio-demographic data, data on their behaviour, personality characteristics, attitudes towards learning and the entire teaching process organization tends to classify students into one of two categories of success. Performance is measured by students’ grade point average achieved over the period of studies. We tested three methods of data mining: logistic regression, decision trees and neural networks. We believe that the presented model would serve as a test for the creation of a broader base of updated data by using some of the information tools and that, based on this model, a number of attributes that would relatively reliably predict the performance in studying will be defined.
Key words: backward stepwise analysis; CART algorithm; decision trees; logistic regression; neural networks.

---

Sažetak
Rad se bavi stvaranjem modela za predviđanje uspješnosti studenata tijekom studiranja primjenom rudarenja podataka (engl. data mining) i analizom čimbenika koji utječu na postignuti stupanj uspješnosti. Model koji je stvoren na temelju socio-demografskih podataka o studentima, podataka o njihovu ponašanju, osobnim karakteristikama, stavovima prema učenju i organizaciji cjelokupnog nastavnog procesa svrstava studente u jednu od dviju kategorija uspješnosti. Uspješnost u studiranju mjeri se srednjom prosječnom ocjenom koju studenti stječu tijekom studiranja. Ispitali smo tri metode rudarenja podataka: logističku regresiju, drvo odlučivanja i neuronske mreže. Smatramo da bi prikazani model mogao poslužiti kao test za stvaranje šire baze ažuriranih podataka korištenjem nekih informacijskih alata i da bi se na temelju toga modela mogli definirati brojni atributi koji bi relativno pouzdano predviđali uspješnost studenata u studiranju.
Ključne riječi: CART algoritam; logistička regresija; neuronske mreže; stablo odlučivanja; stupnjevita analiza unatrag


Full Text: PDF

Refbacks

  • There are currently no refbacks.