TY  - BOOK
AU  - Kargbo,Mustapha Mohamed
AU  - Azyeren,Yasemin Bay
TI  - GENDER PREDICTION FROM INLINE SIGNITURE BIOMETRICS
PY  - 2022///
KW  - Biometry
KW  - Dissertations, Academic
KW  - Machine learning
N1  - Thesis (MSc) - Cyprus International University. Institute of Graduate Studies and Research Management Information Systems Department; Includes bibliography (sheets 48-53)
N2  - ABSTRACT
With the improvement and huge introduction of technological devices and systems, 
the use of biometrics has become a very important inclusion in our lives to increase 
the level of confidence and security with access privileges to approved or certified 
individuals. One of the most popular applications of biometrics is signature biometrics, 
especially in terms of identification and/or verification process; however, it is not 
limited to this, and it reveals high-level discriminative physiological features. 
Although modalities of soft biometrics do provide details about an individual which is 
important for forensics when specified identities cannot be provided, they also include 
non-unique attributes of individuals like gender, age, ethnicity, presence of scars or
tattoos, etc. The purpose of this study is to predict gender from the online signatures 
of an individual applying different feature sets during the classification process. 
In this study, Cyprus International University's handwriting and signature database is 
used for the investigation and prediction of gender from online signature biometrics. 
The database comprises 134 participants’ signatures, providing 8040 signature 
samples in total. The prediction is executed in three fundamental steps; feature 
selection, feature extraction, and classification. The feature selection method using 
WEKA helped to achieve the goal of gaining a better understanding of pattern 
recognition in gender prediction from online signature biometrics and this helped to 
achieve high accuracy levels using different classification methods such as KNN, 
Random Forest, and JRip.
There were 5 experiments conducted in total using different sample sizes and feature 
sets and high gender prediction accuracy rates were obtained in all experiments which 
proves that signature biometrics contain valuable information especially gender and it 
is possible to predict gender from one’s signature. The highest result was obtained in 
Experiment 3 with 97 % accuracy.
Keywords: Biometric, Gender Prediction, Online Signature, Machine Learning
ER  -