Personality Prediction System Based on Signatures Using Machine Learning

Abstract

Many methods are used to assess a person's character and personality. Some are through the face, body movements, body language, handwriting and signatures. Assessing a person's character and personality by looking at the type of handwriting and signature can be learned with the science of graphology. Detecting a person's personality system based on a signature pattern automatically is still difficult. This study aims to build a system of predicting personality based on signature patterns using machine learning. A person's personality based on a signature has many features. In this study the analysed features consist of four features, namely curve start, end streak, middle stroke, and underline. The steps taken in developing a prediction system are model training and model testing. The method used to extract features is Principle Component Analysis (PCA) and the method for classifying is Support Vector Machine (SVM). Based on the test results using confusion matrix produces an accuracy value of 71%. It can be concluded that machine learning can be implemented to predict personality based on signatures with good accuracy.