Thesis (MSc) - Cyprus International University. Institute of Graduate Studies and Research Management Information Systems Department

Includes bibliography (sheets 54-62)

ABSTRACT
Traditionally, biometrics has been associated with fingerprint mapping, facial
recognition, and retina scans. However, the recognition of individuals merely based
on their handwriting has drawn the interest of researchers in the subject of handwriting
biometrics. This new discipline has promising prospects for predicting student
academic performance from their online handwriting biometrics. The available
handwriting biometric database for this research is the CIU handwritten database,
which includes student performance in addition to their identification labels. 804
handwriting samples collected from 134 students are contained in the database.
This research investigates the potential of handwriting biometric features in predicting
academic performance. Two datasets were extracted from the CIU handwritten
database, with the first consisting of 134 participants and the second consisting of 97
participants. Success scores were classified as good, average, or poor in the first
dataset, and as good or poor in the second dataset. NaiveBayes, J48 decision tree,
RandomForest, and KNN models were used to classify the success scores based on
handwriting biometric features. Results showed that the RandomForest model
outperformed the other models in both datasets, with the highest accuracy achieved
after feature selection. The results of the empirical analysis based on the online
handwriting biometrics show that it is possible to predict student academic
performance from their online handwriting biometrics.
Keywords: Biometric Features, Handwriting Biometrics, Performance Prediction,
Students Performance