[3F4-GS-10k] AI応用：行動分析

[3F4-GS-10k-04] A Machine Learning Analysis of the Manifestation of Preference Bias in Behavioral Economics

〇Akira OKABAYASHI¹, Satoshi KAGEYAMA² (1. Kyoto University, 2. Univ. of Tokyo)

Keywords:Machine Learning, Behavioral Economics, LightGBM

In behavioral economics, in the case of wagering, where human cognition intervenes, there is a phenomenon of overestimation due to excessive preference for alternatives with low probability of winning and underestimation of alternatives with high probability of winning (Favorite Longshot Bias).
There has been a lot of interest in research on prediction models that explicitly take into account such biases caused by people's cognition, or preferences, that change from the true evaluation.
The two main methods of analysis are "fundamental analysis" and "technical analysis".
In this study, we used LightGBM as a fundamental analysis to create a model for a mathematical model that has already been designed by technical analysis, and analyzed the preference bias. As specific data, we used data on horse racing odds and their payouts (so-called recovery rates), where the preference bias is pronounced, to create the model and analyze which information the preference bias is pronounced for.