Hidden Markov Models (HMM)

Hidden Markov Model code developed from scratch to be used to identify deceptive behavior from the ROC-HCI deception dataset.

Authors: Taylan Sen and Matt Levin

Main Files/Folders

• hmm.py - Implements and test a hidden Markov model
• truth_bluff.py - Dual-HMM classification on a dataset using cross-validation
• lstm_hmm.py - Modified HMM to avoid exponential decay of prolonged hidden states
• runner.sh - Used to deploy code on a BlueHive computing cluster (SLURM system)
• analysis - Jupyter notebooks to analyze results and code to apply confidence-based thresholds to results
• sh - Shell scripts for dispatching jobs to obtain the dataset and other utilities