Analysis of Value Systems

This chapter describes the use of non-negative matrix factorisation (NMF), an algorithm for the decomposition of data sets into a small number of latent features. Coupled with a model selection mechanism, NMF is an efficient method for identification of latent features and provides a powerful method for the discovery of latent patterns in data. We use convex NMF together with a model selection policy based on consensus clustering to analyse data about the states of agents in the simulation. This can provide insight into relationships between individuals and about the general wellbeing of the simulated population.