Application And Research Of Information Theory In Quantitative Transaction

Due to the complexity,diversity and fickleness of the time series of stock price,there are many factors that affect its change.Some of the factors can be measured,while others are difficult to quantify,calculate or evaluate in a scientific way.To this end,we introduce the theory of communication data analysis methods,and try to use the knowledge of communication theory to quantify the transaction market to explain and modeling.This opens up new ideas for analyzing financial time series.In this paper,based on the communication theory,we combine the current communication theory and the research results of machine learning data mining direction for further study.We divide the process of quantifying the transaction into two parts.One is studying the feature selection of the time series of financial prices,the other is the research on the trend forecast of the time series of financial prices.First of all,as a key technology of data mining and machine learning,feature selection can effectively improve the accuracy of machine learning and the efficiency of the algorithm.However,as data dimensions continue to increase,existing feature selection methods pose serious challenges in terms of effectiveness and efficiency.Maximum correlation is a valid measure of the correlation between variables.For the study of feature selection,we propose an improved network maximum correlation(NMC)model.It can quickly and efficiently calculate the statistical dependencies between feature sets and label variables.In addition,we propose a new NMC-RFE feature selection method based on recursive feature elimination(RFE)algorithm.Second,forecasts based on observed data are one of the main purposes of data analysis,which has a huge impact in the quantitative trading market.Machine learning and data mining are widely used to deal with this problem.In this article,we model and explain this problem based on the Bayesian network model.At the same time,based on the data analysis of communication theory,we simulate the relationship among data variables as a generalized social network.The direct causal relationship from one data variable to another is equivalent to the information transmission through the communication channel.Therefore,predictions based on data variables can make the most of the information transmitted over the communication channel.We use the financial market data to show that the use of this simple method has a good performance.