Application of Truck Flow Big Data in Forecasting Regional Consumption

Abstract:

This study aims to predict consumption changes， which can help precise policymaking， further promote the recovery of consumption. Based on 485 million truck crossing records generated by highway toll stations in Fujian Province， we combed monthly truck flow data from 2017 to 2023 within Fuzhou City， as well as between it and 8 other prefecture-level cities in Fujian Province. Firstly， we summarize the time-series and cross-sectional features to analyze the correlation between truck flow and consumption. And then we use regression model to examine the correlation differences before and after 2020. Furthermore， we select the truck flow indicators with higher correlation in the above analysis， using Ordinary Least Squares （OLS） method and Vector Autoregressive （VAR） model to construct a prediction model of the Total Retail Sales in Fuzhou City. The results indicate that the truck flow and the total consumption of Fuzhou City show a similar trend， growing steadily before 2020 and declining significantly afterward. And the year-on-year percentage of both are highly correlated. The empirical results show that the truck flow from most prefecture-level cities into Fuzhou City is positively correlated with the total retail sales of Fuzhou City， and this correlation is weakened after 2020. Compared with the OLS method， the multivariate VAR model has a higher prediction accuracy with its probability of absolute error less than 1 pct reaching nearly 95%， which allows for the advance prediction of the monthly total consumption in Fuzhou City.