This study aims to find how people’s preferences change over time in public good games by clustering subjects into multiple groups of preference according to their behaviour inside an experimental game. After collecting data based on separate segments of time and clustering each segment individually the difference between any two segments is measured using three different methods firs is considering clusters as groups and subjects as members in these groups then by intersecting these groups we can obtain changes percentage between any two clusters. Second method is by using area under the curve to find the agreement of people inside the same clusters. Third method is by using external cluster validity indices to measure similarities between two groups of clusters. till now the most promising method is area under the curve as it provides simple single number to show differences and it has obvious performance over two other techniques.
Fattah, Polla, Uwe Aickelin, and Christian Wagner. “Clustering Human Behaviour in Public Good Experiment” a First Year Report of PhD (2013).