The Study And Realization Of Shipping Income Prediction Time Based On Time Series And Neural Network

With the development of enterprise informatization and digitization of constant depth, for the analysis of the data requirements are also getting higher and higher. Enterprises for the financial data are no longer satisfied with the requirements of calculation accuracy. With the development of enterprise liquidation cycles continue to shorten, the current focus has been gradually put on the timeliness of financial data on calculation. Aiming at the needs of enterprises, freight income data of enterprises are analyzed and the research forecast method. Due to the actual production process data distribution is not uniform, so using different methods for analysis and processing.This dissertation mainly do the work of three. First, according to the freight income data can be in the form of linear distribution, proposes to use the prediction method based on time series. In this work, in addition to the moving average method described the present freight revenue forecasts are frequently used, also discussed the use of exponential smoothing and linear regression prediction process, and discusses in detail the value reason important parameters used in these methods. Second, for a freight income data curve distribution, using multivariate function regression method, BP neural network prediction method. In this work the exposition, in addition to the prediction given by the calculated value, error of predicted value was analyzed, explained the value reason in this dissertation, BP neural network of each layer neuron number. Third, combined with the front of the two parts of the content, the prediction method using BP neural network and time series analysis, to forecast the freight income data. In this work, in addition to discusses the forecast method, also points out the difference of this prediction with previous methods of forecasting method. Analysis of this method compared to the front of the advantages of the two methods.Through the discussion of this dissertation, for the realization of prediction method in most cases company freight income data. For the software system is developed, the use of commercial software tools to predict, can not be directly transplanted into the system. In this dissertation, according to the actual case of freight income and different data, a detailed analysis of the process of prediction. For each prediction technology, and discusses in detail the specific algorithm. Through the discussion of this dissertation, these techniques can be implemented in the system. For companies, through revenue forecast to adjust the operation condition, it has an important significance for the management of the company.