Rationale Classification for Educational Trading Platforms

Annie Ying, Pablo Duboue

Abstract: Stock market trading simulation platforms have become popular finance education tools in recent years. To encourage students to think through a trade order, many of such platforms provide a field called “rationale” in the trade order user interface.In this paper, we first present a novel problem called “thoughtful rationale classification” based on two studies: (1) an observational study on the factors affecting a finance professional assessment of a student’s trading sophistication and (2) a qualitative study on 2,622 rationales. The two studies together reveal that when a student provides thoughtful rationales, defined as rationales that document external research, specific strategies, or any technical analysis performed, the student is likely to be assessed higher in terms of the trading sophistication. We show that labeling rationales as thoughtful or not is a well defined task and automate it using CNNs. We also compare baseline implementations using simple features and support vector ma-chines over selected keywords.