Empirical Study on Intercity Logistics Distribution Demand Forecast Based on Grey-Markov Model

Abstract

The forecast of logistics demand is the basis of decision-making to set down logistics programming and build logistics base. In this paper, a Grey Markov Model for forecasting logistics demand is presented by means of combining Grey system theory with dispersed Markov Chains theory. Firstly, according to the variable sequence, a GM model is setup to forecast the general development trend of variable as first fitting values. Secondly, the fitting values are revised by means of Markov state change probability matrix for forecast as the second fitting values. The model overcomes the influence of random fluctuation data on forecasting precision and widens the application scope of the grey forecasting. In this paper, empirical studies of logistics demand forecast were performed. Results show that the Grey Markov Model has higher accuracy than that of GM model. Logistics demand forecast is one of the core areas of the logistics system planning. It is the foundation basis to determine the size of logistics systems and management models. Logistics system is a complex socio-economic system and it is affected synthetically by social, economic, physical and other factors. Thus, logistics system has larger random, fuzzy and gray features. These features make the forecasting difficult. Gray prediction and Markov chain prediction can both be used for sequence prediction. However, the geometric curve of gray prediction has monotonically increasing or decreasing trends, so it is difficult to reflect greater volatility in the original sequence and it will only reflect the overall trend. Markov prediction object is randomly changing dynamic system. Its projections are based on transition probability between states to speculate the future development of the system. Transition probabilities reflect the degree of influence of various random factors. So, Markov chain is suitable for volatile random sequence prediction problem. This just can make up for the limitations of gray prediction. Therefore, these two methods have highly complementary features. However, if they are used in combination, higher prediction accuracy may be obtained. But the Markov process requires no after-effect of the state. If we adopt the GM (1, 1) model to fit time-series data of prediction, then by identifying the trends in their residuals we can not only make up for the limitations of Markov prediction but also compensate for deficiencies in the gray model. According to the features of logistics system planning encountered in the actual situation, a gray Markov chain model was put forward to predict the demand of logistics distribution.