Application of bayesian poisson tensor factorization for moving home simulation analysis

Abstract

Today, by the benefit of globalization, population movements have become much more liberal than before and have played a major role in bringing about socioeconomic changes. We implemented the migration of residence as a representative population movement in the modern country as an agent-based model, and we subdivided and analyzed the simulation results by machine learning method. The moving home simulation is based on the mixture of Schelling's segregation model and Urban Suite model. And we introduce Bayesian Poisson tensor factorization (BPTF) to decompose the agent movement information, which is a 5-way tensor and very sparse. There are simulation results and BPTF results. Under the sudden large change of the input parameter, we could observe that the agents move toward their own gains. And BPTF catches the change time intervals. A more advanced model will help to establish a housing policy by mimicking the real world and detecting more hidden information in advance.