Sales Forecast For Regional Automotive Market Using Adaptive Network-based Fuzzy Inference System

Automobile industry is one of the important manufacturing industries in our country, and has high linkage with other industry, so that it can lead related industry to upgrade. After the participation of WTO, automobiles manufactures in our country will face more aggressive competition. How to handle the market trend and forecast the sales with high accuracy is the key success factor. Sale forecasting of automobile market is the data basis of planning and control for the departments of automobile company such as production, logistics and sales. Improving the sales forecasting accuracy has become a primary concern for automobile company.Existing prediction method taken no account of the characteristics of sale forecast process, regional market and the distribution of different vehicle segment, so the prediction results are useless to automobile company's production planning; Existing prediction method taken no account of the change and choice of influence factors which have enormous implications to prediction quality. Thus, to construct the prediction frame of the regional automotive market, including complete explanatory variables in order to enhance the predict method that make up the deficiency previous prediction methods is the major motivation of this research.This paper combines regression analysis method and fuzzy neural network method and puts forward a sales forecast model for regional automotive market using Adaptive Network-based Fuzzy Inference System (ANFIS). On condition that discriminate the influencing factors of automobile sales and considering the feature of regional market, we developed a sales forecasting method that considers 14 variables such as GDP,CPI, oil price, automobile sales price, sale leads and advertisement. First, we use the principal component analysis (PCA) and partial least-squares (PLS) method to select most influential variables as our input variables. Then, we input the influential variables and sales in ANFIS to obtain the forecast. Finally, this paper compare the predictive results with two forecast models with examples: autoregressive integrated moving average model (ARIMA) and time series decomposition method. We compare the three predict values to actual sales data and use Mean Absolute Percentage Error (MAPE) as the evaluation of prediction accuracy. Experiments results show that the ANFIS model that consider change and choice of influence factors can predict regional automotive market sales and has higher prediction accuracy than the other two traditional methods, especially in the new market environment characteristics such as high fuel prices and automotive purchases duty reform. So the presented model provides an important reference for making targeted improvements of sales forecast process by automobile manufacturers and developing sales strategy that adapt to the regional automotive market characteristics.