Investigation of causal public opinion indexes for price fluctuation in vegetable marketing.

Though it is believed that public opinion helps to predict vegetable prices, an intelligent scheme for predicting vegetable prices by online public opinion is missing. This research proposes an intelligent vegetable price prediction model, TPCDR, which includes five modules, i.e., topic matching, public opinion computing, causing index selection, de-noise, and regression of price prediction. To investigate the correlation between vegetable price and public opinion in a fine-grained manner, this research considers four types of economic causing indexes, i.e., natural environment, supply, demand, and economic policy. TPCDR model integrates Granger causal testing and linear regression and discovers the significant causal factors raised from the natural environment and supply. Meanwhile, through vegetable price forecasting experiments, our model outperforms other baseline methods. This result suggests that the TPCDR pipeline is reasonable and reliable to model the association between public opinion and vegetable price. • An intelligent model for the association between public opinion and vegetable price. • Four types of economic causing indexes considered. Causal relation obtained. • Constructing a corpus for online public opinion with vegetable prices in China. [ABSTRACT FROM AUTHOR]