I was the mentor of a student from Polytech Nice-Sophia, working on a project aimed at sing machine learning to understand economic development, called “Using machine learning to understand economic development”.
Economies grow by updgrading the products they produce and export. Recently, the notion of product space has succesfully managed to explain the industrial development of countries by its export basket. The product space is a network of relatedness between products. It connects industrial product by its likelihood of being both exported by the same country. The technology, capital, institutions, and skills needed to make newer products are more easily adapted from some products than from others. Empirically, countries move through the product space by developing goods close to those they currently produce. It is known that more-sophisticated products are located in a densely connected core whereas less-sophisticated products occupy a less-connected periphery. Economies tend to percolate through this network, to increase its fitness. We propose a new method to study the dynamics of economic development, based on a PDE formulation of the dynamics of export, that will be refined by combination with real economic data. You will utilize an approximate physical formulation of the growth of products and its diffusion through the product space, and fit it to data observed from the last 50 years. This should be sufficient to obtain a model which is directly interpretable and accurately describes the economic development process. More specifically, you will make use of a generalized reaction diffusion PDE to try to predict economic development from historical data. The dynamics will be approximated by a Universal PDE composed of two neural networks. One will represent the local growth term, while the second will represent the diffusion (or percolation) operator. The data consists of exported products in many countries over the last 50 years. You will try to apply the model to as many countries as possible, and evaluate its accuracy. Benefits of this work are immense. If we manage to obtain an accurate model, we would provide humanity with a new tool to guide economic policies towards an optimal economic development. If the model is not working as expected, we would anyway provide some insights in the peculiarity of the dynamics of economic development.