CatBoost and Logistic Regression as Machine Learning Approaches in Matchmaking and Perceived Availability

Abstract

This paper aims to redesign the analysis of the “Speed Dating” dataset, which was part of the research titled “Gender Differences in Mate Selection: Evidence from a Speed Dating Experiment,” presented by Raymond Fisman, Sheena Iyengar, Emir Kamenica, and Itamar Simonson in The Quarterly Journal of Economics, the oldest professional journal of economics in the English language, in 2006. Based on the theory of "perceived availability," which suggests that people are more likely to find those who seem more attainable or interested in them to be attractive, logistic regression and the CatBoost ensemble method were employed to uncover patterns that appear influential in the decisions of individuals of the opposite sex regarding the potential for a future relationship from a four-minute speed dating social experiment. The findings indicate that, in general, individuals prioritize the following in their potential partners, from most to least important: attractiveness, perceived compatibility, shared interests, sense of humor, ambition, satisfaction with acquaintances (indicative of sociability), TV interests, sincerity, and partner's age. These results report an accuracy of over 80% with Logistic Regression and 88% with the CatBoost ensemble method. The tool used in model development was Orange Data Mining 3.37.