A novel GDP prediction technique based on transfer learning using CO2 emission dataset. (1st November 2019)

• Record Type: Journal Article
• Authors: Kumar, Sandeep; Muhuri, Pranab K.
• Is Part Of: Applied energy. Volume 253(2019)
• Publication Date: 2019-11-01
• Subjects: GDP prediction -- Transfer learning -- CO2 emission data sets -- Developing economies -- Developed economies; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2019.113476
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: A transfer learning based novel GDP prediction method using CO2 emission data. It can estimate GDP even when macroeconomic data of a nation is unavailable. Simulations consider CO2 emission data of four economically heterogeneous nations. Results are comparatively presented employing three well-known regression methods. Capable of predicting the missing GDP values of isolated and war-torn countries. Abstract: In the last 150 years, CO2 concentration in the atmosphere has increased from 280 parts per million to 400 parts per million. This has caused an increase in the average global temperatures by nearly 0.7 °C due to the greenhouse effect. However, the most prosperous states are the highest emitters of greenhouse gases (especially CO2 ). This indicates a strong relationship between gaseous emissions and the gross domestic product (GDP) of the states. Such a relationship is highly volatile and nonlinear due to its dependence on the technological advancements and constantly changing domestic and international regulatory policies and relations. To analyse such vastly nonlinear relationships, soft computing techniques has been quite effective as they can predict a compact solution for multi-variable parameters without any explicit insight into the internal system functionalities. This paper reports a novel transfer learning based approach for GDP prediction, which we have termed as 'Domain Adapted Transfer Learning for GDP Prediction. In the proposed approach per capita GDP of different nations is predicted using their CO2 emissions via a model trained on the data of any developed or developing economy. Results are comparatively presented considering three well-known regression methods such as Generalized Regression Neural Network, Extreme Learning Machine and Support Vector Regression. Then the proposed approach is used to reliably estimate the missing per capita GDP of some of the war-torn and isolated countries.