Use of WASD neuronet in projecting the population of Oceania based on 1000-year historical data

摘要

With growing economic and political influence in the world, the important role played by Oceania in population issues should not be neglected. So it is very important and urgent to find an effective way to make a proper population projections. Nonetheless, the traditional methods focused on fertility and mortality may lead to the lack of all-sidedness in projection results. We realize that the historical data contain the internal mechanism of the population development, and the neuronet performs well with nonlinear data and the multifactor system. Therefore, in this report, we construct a 3-layer feed-forward neuronet equipped with a weights-and-structure-determination (WASD) algorithm to learn the historical data and project the population. With the neuronet well trained by over 1000-year historical data, we successfully project that the future Oceania population will keep a steady increasing trend in the coming fifteen years.