Supervised Learning Approaches for Deceit Identification: Exploring EEG as a Non-invasive Technique

Deceit identification has been a problem since ages and for generations past now. The process is taken in both positive and negative prespectives. Positive for the justice it delivers to the unjustified people and negative for the techniques that are used for the purpose. It has been historically proved that the techniques involved in lie detection (common term used for deceit identification) involve many methodologies including those against human rights and various international conventions. The processes involved have always been in a questions and various studies have tried to prove the deceit; hence, latest methodologies and developments in machine learning area are under trial for such detection jobs. The following paper discusses various supervised learning methodologies like k-Nearest Neighbours, AdaBoost, etc., in order to prove deceit. As general awareness the deceit information used for supervised learning is already labelled and is taken from an open-source dataset. The study tries to establish a basic threshold of parameters in lie detection (LD) and is comparable to the human levels of information gathering with deceit identification with an accuracy of over 70% which is comparable to that of human interrogation techniques. The paper aims to lay a basis for deceit detection using machine learning methodologies which in the future could change the face of LD.