An Econometric Model of GDP using Machine Learning through an Artificial Neural Network

Abstract

Forecasting economic strength from several economic indicators is the primary concern of the area of econometrics. Gross domestic product measures the current value of all goods within a country and is one of the most prominent economic parameters used to evaluate economic development. With the understanding of how economic indicators affect the economy, planners can choose to allocate resources in certain industries to boost economic growth. This study builds an econometric model of US GDP with back-propagation artificial neural network architecture. Taking data from more than the past 70 years, this model will use machine learning to understand the behavior of three economic sectors of the United States. By analyzing the interconnecting behavior of the three sectors, it can form a mathematical relationship between them and produce an output of US GDP. This model accounts for nonlinearity and can be easily adapted to include more economic indicators than just the three sectors selected, sharpening the results and allowing the study of new relationships. After the machine learning process is complete, economists can adjust the input values to examine its effect on the resulting GDP from the derived mathematical relationship. The model's current implementation is found to be very satisfactory and can be useful for the future planning of economic activity.