Forecasting model for a rigid set of samples

This paper deals with a modeling method for dynamic forecast of a rigid set of samples in a multi-dimensional space. This method can keep the relations of the sample points to maintain the entirety of the set as well as possible. What's more, only the independent factors of this rigid set will be at work by using stochastic models. In this way, a large amount of modeling work can be cut down, especially when the number of sample points in the rigid set is very large. At the same time the precision and reliability of the forecast can be better ensured.