In China, vegetable production plays a critical role directly to national economics and social stability. With the rapid development of social media, public opinions through the Internet is transmitted to the vegetable market in a direct way Previously, focus of most systems was to investigate whether large scale network public opinion is capable of affecting or predicting vegetable price changes. In this paper, we analysed the impact of network public opinions based on a hybrid research strategy. The strategy combined natural language processing (NLP), convolutional neural network (CNN) and classic economic methods. First, we designed corpora indicating different domains of the vegetable market, including supply, demand, natural environment, and government policy; second, we used CNN to perform the topic modelling upon public opinions in a large scale; Third, we investigated vegetable prices volatility with Granger causality test on account of time-lag effects, and results show the correlation between vegetable prices and three public opinion indicators, i.e. demand, supply and natural environment. A subsequent multiple linear regression model augmented the results. Eventually, two linear and three nonlinear predictive models were presented. Results suggested that there is a weak linear correlation between vegetable prices and these three public opinion indicators. This indicated that network public opinions have a potential impact on vegetable prices volatility and can be treated as potential factor to predict vegetable prices.