Fingerprint classification system for online shopping mall using Machine Learning tools

Abstract

Internet shopping has become more popular and some customers are still reluctant to purchase products online due to the perceived risk. The minutia features of biometric fingerprint do cause a major barrier in existing biometric recognition system. The use of Machine Learning (ML) technique with the help of a hash function to provide a better and stronger security system for online customers. The unique features of fingerprint data were converted to a unique numerical code and helped prevent all cases of breaches, hacking and other intrusive actions experienced with existing online shopping platforms. The Artificial Neural Network (ANN) and logistic techniques were employed to train and extract the minutiae properties of ridge-Island, rending, dot, enclosure and Bifurcation for developing a secured fingerprint image recognition system. A hash function that converts minutiae features of fingerprint biometric data to form a unique numerical code was introduced. The ANN and Linear Regression (LR) machine learning techniques were employed to improve upon the existing methods of human biometric fingerprint recognition system that provides uniqueness to different users. The ANN model recorded 96% recognition accuracy which was higher compared to the Logistic regression that produced 93% metrics of accuracy; also, the ANN technique produced 1.288410 error value measured to be smaller than Logistic regression with 1.469694 error rate.