Design And Implementation Of Market Demand Forecast Model For Automobile Marketing

Automobile industry is a pillar industry in our country. Nowadays the domestic auto industry has entered a high-speed growth period after the development of the initial stage. However, a direct sequence of rising in vehicle production is fierce competition, especially after China joined the WTO, with the pressure that famous international automobile manufacturers have entered our country, the second largest auto market all around the world, the competition of domestic auto industry has become extremely intense. Then how to make a decision on business operations correctly and timely for the fast-changing market conditions becomes an important topic to all the managers of automobile enterprise.Before a significant decision about the enterprise is made the leader must have an understanding deep inside the industry market, which means not only knowing the history of the market and the analyzing the current status, but also giving the prediction of the future trends. Thus, the correct market forecast is the basic for the correct implementation of enterprise marketing and sales management; still it is the prerequisite for the enterprise to make decision.Market forecast is the prediction of the further market based on the data of history status, so that large amount of history data is needed. As the rapid development of enterprise informationization construction all over the world, more and more domestic automotive enterprises have carride out application systems such as ERP, CRM, SCM and so on, and formed their own basic database gradually. The data stored in the database can be the high-quality source for the market forecast. Accuracy forecasting the future market demands with full usage of the data source statistical methods and data mining can help the enterprise managers greatly. Based on this purpose, we developed the auto market forecast system.This system is the sub-system of"Intelligent Decision Support System for the Automotive Marketing"developed by JiLin University. The intelligence decision supporting system, based on business intelligence technology, is a large-scale software which realize the intelligent decision on the process of automotive marketing. Market demand forecast is an important function of the decision supporting system. The sub-system can make a quantitative prediction for the future trend of the automobile market using BI technology.The auto market demand forecast system can provides a general function of the auto market forecast for the users, containing data selection, model creation, prediction implementation, result presentation and etc. Simultaneously, the system has low complexity to operate, and is designd using a humanization way of thinking. It also displays a friendly interface to the users. The users can operate the system freely according to their actual needs by just learning a little of the system.The system can be divided into three layers by the architecture. Data layer provides data to the forecast function and stores the data generated by the system. Function module layer can be subdivided into business operation layer and technical implementation layer. Business operation layer organizes algorithms and data to realize the business process according to the specific requirements of the users. Technical implementation layer provides various types of algorithms to predict. Application layer is user-oriented and provides graphical user interface to make the user and the system interact with each other, users submit a specific operation order on the application layer, then the system executes it and display the result in various forms such as tables, graphics and numbers.The system can be divided into three function modules: data management module, model management module and prediction implementation module. Data management module is in charge of generating various types of data for the system, including the data for training model and symbolizing different kinds of prediction objects. Simultaneously, the data management module provides functions such as selection, addition, deletion and modification for the users to operate the data. Model management module takes the creation, update and deletion of the prediction model in charge. Here the model means the predictor obtained by training of the history data. Users must construct their own prediction model they need before they make actual prediction. After the model creation, users can look up the model, update the model or delete the model optionally. Prediction implementation module accomplishes the prediction acquirement of the users and displays the prediction result to the users intuitively.Using only one algorithm to get the results of the market prediction would be biased, as a wide range of history data will shows the characteristics of a complex and changeable, different types of markets require different prediction methods. So the system provides a variety of prediction algorithms for the users. By comparative studying of history data, we choose four prediction algorithms for the users, including linear regression method, exponential smoothing method, grey system method and BP-ANN method. These algorithms have their own advantages and disadvantages for different types of history data and prediction demands.In this paper, with the implementation of these methods, simultaneously we study the characteristics and applicability of each method in the field of automotive market prediction by the result these algorithms applied in accordance with the actual prediction of the automobile market.Linear regression set up the regression equation on the basis of analyzing the relationship between the variables and predicts the dependent variable value by the independent variables. The system realized unitary and multiple linear regression methods both. For the history data that include several variables, the system will set up several models, then compare the accuracy of the models and select the best model automatically for the users. In view of the trend of linear growth in China's auto market in the past years, linear regression method would be quite suitable for the prediction of medium and long-term market.Exponential smoothing method takes some mathematical methods to eliminate the numerical level of mutation of the time series, and then obtains a trend sequence and evaluates it. Based on the single exponential smoothing, our system realized quadratic exponential smoothing, which solves the problems that evident trend sequence can't be predicted in single exponential smoothing. Exponential smoothing prediction method fits the short-term market for its easy calculation and obvious lag.Grey system forecast method set up the differential equations using less or inaccurate original data sequence, and is usually used in time sequence prediction. Grey system theory researches and predicts the unknown field through the known information to realizes the whole system. Grey system forecast method will receive a good performance when there are not enough history data.BP-ANN method (Back-Propagation Artificial Neural Network), process the training data set iteratively, and compare the prediction value of the network and the actual value for each sample, then modifying the network weights constantly through the error comparison to obtain the accurate model. BP-ANN method suits the long-term auto market prediction for its high prediction accuracy and strong capability of errors.To sum up, we designed and developed a market forecast system for the managers of auto industry, which implements 4 classic prediction algorithms, and then we discussed the scope of applications of these algorithms. The system also places a lot of room for improvement, and we will consummate unceasingly this system in our further work.