An exploratory study on promising cues in deception detection and application of decision tree

Five comparative decision tree experiments demonstrate that tree performance, such as accurate rate and complexity, is dramatically improved by statistically and semantically selecting cues.

Abstract

Automatic deception detection (ADD) becomes more and more important. ADD can be facilitated with the development of data mining techniques. In the paper we focus on decision tree to automatic classify deceptions. The major question is how to select experiment data (input data for training in decision tree) so that it maximally benefits the decision tree performance. We investigate promising level of the cues of experiment data, and then adjust the applications in decision tree accordingly. Five comparative decision tree experiments demonstrate that tree performance, such as accurate rate and complexity, is dramatically improved by statistically and semantically selecting cues.