DOCUMENT VERIFICATION SYSTEM FOR FRAUD DETECTION USING MACHINE LEARNING TECHNIQUE

Abstract

This research paper discusses an intelligent document verification system for fraud detection using machine learning technique. This research seeks to address the problems encountered in the conventional verification system such as delay time, lack of intelligence and not being reliable by developing a machine learning based verification system and localizing it for the verification of certificate at the Nnamdi Azikiwe University (Unizik), Awka, Nigeria. To achieve this, the methods of data collection, data acquisition, data processing, feature extraction, artificial neural network training, and classification were used. Self-defining equations and modeling diagrams were used to develop the artificial neural network model and then train with 1180 authorized data collection of Nnamdi Azikiwe University, Awka certificates from 2016 to 2020, to generate the reference verification model which was used to develop the expert system for verification of documents. The system was implemented using image acquisition toolbox, image processing toolbox, statistical and feature extraction toolbox, neural network toolbox in Matlab and then tested for evaluation. The result recorded however, achieved a Mean Square Error (MSE) performance of 0.000100Mu and Regression value of R= 0.99373 which is very good, with implication that the proposed system is very reliable.