Exploration of Feature Selection and Advanced Classification Models for High-Stakes Deception Detection

Recent research has demonstrated the effectiveness of automated text-based deception detection. In this study, using a variety of data sets and common classification techniques, this has been shown to be an accurate technique. Previous results have shown the need to reduce the number of inputs to these models in order to prevent overfitting. While previous results have been promising, there is a need to improve accuracy and reduce the number of false positives. Using 5 classification models and 3 variable sets, we have achieved accuracy level of 76% in this study.