Unbiased Grey Markov’s Optimization Modeland Its Application To The Forecast Of Commodity Sales

Since China’s accession to WTO, Most of the domestic companies will participate in the international competition in the market and continue to receive goods from other countries in the world in our local market dumping pressure.Therefore, doing well in commodity sales forecasting is the key to making the right business strategy and marketing plans. Scientific and rational commodity sales forecasting model has a very important significance for enterprise to make the right marketing plan.So far, many scholars and researchers have made a lot of work in terms of merchandise sales forecast. Such as Markov model, gray prediction model and neural network model to predict sales of goods. But mostly of the model is a single model,the prediction accuracy is not high. Firstly, this article used gray system theory to establish unbiased gray GM(1, 1) forecasting model which basing on the traditional gray forecasting model, eliminating the inherent bias of the original gray prediction model and improving the prediction model’s anti-jamming capability. Secondly, this article used the theory knowledge of Markov chains to correct unbiased GM(1, 1)model’s relative residuals, establishing unbiased GM(1, 1)- Markov model. Both use the model to predict the development trend of the data series, but also to reflect the volatility characteristics of the data. Finally, the article used particle swarm optimization Algorithm to albino Unbiased GM(1, 1)- Markov model gray interval parameters, obtaining Unbiased GM(1, 1)- Markov model after PSO’s Optimization,and prediction accuracy of the model has been significantly improved. The paper selected cases from 2007 to 2014 1-9 month US light vehicle manufacturers in Asia division of Subaru car’s cumulative sales. Experimental results show that the optimized by PSO Unbiased GM(1, 1)- Markov model’s predictions mostly according the realistic car sales. The prediction model’s accuracy is significantly higher than Unbiased GM(1, 1)- Markov prediction’s accuracy. Thus, the model can be used to predict the actual sales of goods, and thus providing the basis for corporate’s business decisions.