Predicting vegetable prices based on socio-economic indicators using social media

Abstract In China, vegetable production plays a particularly critical role directly proportional to national economic and social stability. With the rapid development of social media, network public opinions can be transmitted to the vegetable market through the Internet easily and quickly. Before, most systems attempt to investigate whether network public opinion, as expressed in large-scale collections of news about vegetables, is capable of affecting and predicting vegetable price changes. Although the links between these are obvious, the analyses highlight the perfectibility of the research conducted in predicting vegetable prices and the necessity of systematically pursuing them. This paper analyses the impact of network public opinions based on Natural language processing (NLP) and Convolutional Neural Network (CNN), derived from designed economic factor corpora for different domains around the vegetable market on vegetable prices volatility using Granger test and the multiple linear regression model. Empirical results suggest there is a linear correlation between vegetable prices and network public opinions for three domains of natural environment, demand and supply. And network public opinions for economic policy have a short-term influence on vegetable price volatility. This indicates that network public opinions have a direct impact on vegetable prices volatility by influencing the expectation of market traders, and to some extent, they could be adopted to predict vegetable prices.