Forecast And Visual Analysis Of Methanol Product Price Based On Multi-source Data

With its rapid development of economy, China has developed into an industry power in a short period of time. Methanol plays an important role in the industry development of China. The price of chemical products is time-varying, instability and volatility, which influenced by many factors, such as seasonal factors, monthly output, macroeconomic indicators, etc, and it is paid attention by domestic and international. The traditional prediction methods based on expert experience and statistic methods, which is difficult to predict the prices accurately and effectively. The commodities are about national strategic interests, the rise and declining of the whole nation economy, more and more experts join in the prediction of chemical products’price.The prediction of multidimensional factors, we screen these factors according to the knowledge of statistics as well as the experience of experts, at the same time propose the forecasting model based on multi-source which combine network emotion value with expert experience. Firstly, we deal with the history data and analyze the correlation of them, choose the appropriate models to forecast the data and compare the forecasting errors, and establish the optimal long term forecasting model which combine GARCH and ARMA. Secondly, through mining and processing, statistics and analysis of the methanol products’ network data, we can construct emotion dictionary for it and obtain the emotion value. We design special expert questionnaires to experts in different industries for a long time, and achieve expert experience from quantitative analysis. Lastly, we simulate the new prediction model which combines the emotion value with the expert experience. We also evaluate the prediction errors of both existing prediction models and our prediction model of short-term and long-term models.Based on the forecast of multi-source data, our article designs a visual analytic system with strong interaction and hierarchical structure for methanol data further. As the price of methanol is influenced by multidimensional factors, our system uses time serious diagram to reflect the historical price and emotion trend of methanol, while display prediction results of multidimensional data effectively; The prediction errors is presented by dynamic pie chart; The historical emotional text is searched by search box; At the same time, the experts can infuse their emotion value and viewpoint into the system, and to change prediction result through quantizing the data dynamic and show the final predict result. Through time series graph, interaction and other techniques, our system solves the prediction and display problem of the data. Through the experimental analysis and user tracking, we find that our method has a higher accuracy and our system has more strong practicability.