A New Method Framework For Complex System Analysis And Simulation

The creation and development of system science has led people to realize that the system is hierarchical and functional,and that the system and its environment have continuous exchange of matter,energy and information.At the same time,the system is in constant development and change(evolution),the system is larger than the amount of its components.It is further found that the system of certainty has its internal randomness,while the system of randomness has its own internal certainty,and the system can be stable in a state far from equilibrium.Against this backdrop,Complex Science,the science that focuses on complexity and complex systems,has come into being.The research of system science has entered a new era.In general,complex systems lack simple generalization features,especially when they have some degree of intelligent autonomy.These characteristics of complex systems have an impact on the relevant theoretical analysis methods,applied technology research,and even philosophical thinking.The research work on complex system modeling,evolution analysis and so on is challenging.Benfit from the rapid development of computing technology,scientific computing is the third methodology in which people know the world after scientific experiments and scientific theories.Some complex systems that are difficult to carry out experimental research or theoretical analysis may be studied and simulated by scientific calculation and visualization.At present,there are many modeling methods to explore complex system,each with its own characteristics and different aspects of focus,in general,these model methods can be broadly divided into two categories.The first one is the numerical method such as artificial neural network,cellular automata,genetic algorithm,etc.It is to represent the preference for "fitting" facts.The second one is the statistical method represented by probability(graph),Petri network,complex network,fractal theory and so on.The first type of method is close to the facts(data),high precision and good fitting effect,but there are problems that explain the causes and analyze the structure of the problem.The second kind of method often needs to adopt different research framework for different types of problems,and lacks the background of more unified research paradigm,research framework and corresponding philosophical thinking.Based on the comparative study of the characteristics of artificial neural network,grey system,genetic algorithm,cellular automata,probability(graph)and other complex system modeling methods,this thesis discusses the key points and difficulties of complex system modeling from the background of methodology and philosophy,and proposes a CUP modeling method combined with the characteristics of different complex system research methods.The CUP frame model method aims to integrate the characteristics of probability and statistics and numerical fitting methods,and try to have a good performance in the predictability and interpretability of the algorithm.In this thesis,the theory and application of CUP frame model method are systematically studied.Firstly,this thesis defines the primitive of CUP model method,discusses the CUP structures of different levels and layers,discusses the fractal characteristics and parallel characteristics of multi-layer CUP structure,makes clear the compatibility of the model method with probability based model method and numerical fitting method,and puts forward the standardization method.Secondly,this thesis proposes a coefficient fitting algorithm based on input-output sequence.Combined with the fitting algorithm,the analysis and prediction method of state coding based on CUP is given.Thirdly,the theory and algorithm of CUP model proposed in this thesis can explain the simulation results of some complex systems in a variety of ways because it takes into account both the probability method interpretation and numerical fitting results.Fourth,this thesis also discusses the CUPization method of practical problems through several examples,and makes experimental research on the modeling and analysis process of complex system based on the CUP theory and method,involving the application of artificial financial market,emotion simulation,infection diffusion simulation,sequence analysis,pattern formation and other complex systems.The above experimental results also show that the CUP model has certain universality.The computing architecture corresponding to CUP model algorithm is easy to deploy in homogeneous or heterogeneous computing environment,and supports parallel processing.This thesis has the following innovations:1.A CUP theoretical modeling method for complex system modeling and simulation is proposed originally.The basic concept,structure and properties of the CUP modeling method are studied.This method combines the advantages of numerical fitting and probability induction,and has good interpretability.And on the basis of the proposed theory,the relevant algorithms such as fitting,prediction and analysis are derived.2.Using the theory and method of CUP model,the modeling,simulation,analysis and experimental research of several typical complex systems in natural science and social science are carried out.The results show that the CUP model method has certain universality.3.Design and implement the computing architecture software corresponding to the CUP model method,which is easy to deploy in homogeneous or heterogeneous computing environment and has good usability.