Research On The Forecasting Method Of China’s Corn Futures Market By Combining Sentiment Analysis

The volatility of China’s corn futures prices has intensified in recent years,influenced by policy regulation and multiple supply and demand factors.Since 2016,the corn policy has transitioned from the temporary reserve system to the "price-subsidy separation" system,causing a significant fluctuation in corn prices and leading to a rapid increase in the demand for hedging investments from enterprises and investors.With the development of internet technology,traders are easily influenced by online public opinions.Investors pay attention to various information platforms as a basis for decisionmaking.By classifying the sentiment polarity of online public opinion text data,market sentiment and potential trends can be captured,thereby improving prediction accuracy and helping investors and enterprises formulate wise trading strategies.This thesis starts from the information contained in China’s corn futures market,comprehensively considering multiple dimensions such as demand and supply,and profoundly analyzes the inherent changing laws of the corn futures market.Meanwhile,this research combines public opinion analysis and utilizes existing data to construct a combined prediction model for training,which can be applied to corn futures market price prediction.Focusing on these research contents,we mainly carry out the following three aspects of work:1)Constructing a corn futures market index system combined with public opinion analysis.First,this research analyzes the influencing factors of corn futures prices from multiple dimensions,selecting corresponding proxy variables and sentiment variables of public opinion as input features for subsequent deep-learning prediction models.Next,a parallel dual-channel deep learning hybrid model PDDM is constructed,and its effectiveness in public opinion sentiment analysis tasks is verified through experiments.Subsequently,we introduce a Chinese financial sentiment dictionary and construct the PDDM-CFSD model to further improve the model’s sentiment classification effect on financial domain text.Finally,using the classification results of the PDDM-CFSD model and the moving average method,we construct sentiment index features to verify the correlation and causality between the sentiment index and corn futures prices.2)Establishing a combined prediction model to predict and analyze corn futures prices within a period.Based on the construction of the corn futures market index system combined with public opinion analysis,this thesis first uses classic prediction models such as LSTM and ARIMA for comparative experiments.The experimental results show that different types of prediction models have different characteristics in short,medium,and long-term predictions and verify the effectiveness of the influencing factors and public opinion analysis proposed in this thesis for predictions.The research then combines the characteristics of deep learning networks to construct a corn futures market prediction model based on CNN-Bi LSTM-AM and verifies its superiority through a series of comparative experiments.Finally,we propose a weighted combined model based on different prediction days,using the advantages of traditional time series prediction models and deep neural network models in different prediction days,resulting in good performance of the combined model throughout the entire prediction period.3)Developing a corn futures market price prediction visualization system based on public opinion analysis.This thesis conducts in-depth research on functional and business requirements and designs the system architecture and technology selection.Based on this,we develop a corn futures market price prediction visualization system based on public opinion analysis.