Machine Learning And "Keqiang Index" On Forecasting China's GDP Growth

Since its economic reform,China has experienced extraordinary growth and has become the second biggest contributors of the world economy.However,the middle kingdom is not a mature economy yet and is still growing at a rapid pace.It is,therefore,essential to pay a close attention to its growth tendency,in order to monitor the global economic situation.In this paper,I analyze the forecasting performance of different methods,using the three indicators of the "Keqiang index"(electricity consumption,railway cargo volume and bank loans)to find a suitable model for predicting China,s GDP growth efficiently.To carry out the analysis,I use both national and provincial data from 1995 to 2018;and compare various Machine Learning(ML)algorithms,traditional metihods and the traditional computation of the,Keqiang index"(KQ).To assess the performance of each model,I use k-fold cross-validation,widely used in ML to calculate prediction error.The results show that ML methods outperform the traditional computing method of KQ.In fact,the validation error of all the ML models is relatively low(the lower the better),particularly regression splines.This confirms the assumption that while it shows to be a good altermative measure for economic growth of China,KQ is not a suitable method for prediction.Moreover,the tentative forecasting model created using regression splines demonstrates a better performance.Although it has the same trend,it appears to not have the exact same rate.It is possibly due to the small size of the data or it suggests that there should be more factors to be taken into consideration,This initiates a future issue to study.