Research And Implementation Of Cause And Effect Diagram Structure Learning And Visual Analysis System For Dynamic Network

In recent years,the research and application of causal inference has involved many research fields such as sociology,medicine,and economics,providing solid theoretical support for the arrival of the era of strong artificial intelligence.Causal structure learning is a subtask in the field of causal inference,which refers to learning the causal topology between variables by analyzing observational data.This topology is usually represented by a directed acyclic graph.Although the existing directed acyclic graph learning methods have achieved great success,no researcher has conducted research on causal graph learning methods that combine time-series information and correlation information between samples,and such data is widely available in the real world.In addition,there is no visual analysis system specially developed for the causal graph structure learning algorithm,which can intuitively display the causal graph.Therefore,in view of the above problems,the main research contents of this topic are as follows:Research on causal graph learning algorithms for dynamic networks.This topic firstly models the causal graph learning problem for dynamic networks,and for this problem,researches and develops the causal graph learning algorithm GraphNOTEARS,which is used to learn the structure and parameters of sparse causal graphs from dynamic networks with highdimensional features.In addition,this subject designs and conducts a lot of experiments on both simulated and real datasets.Experimental results show that the performance of our model is significantly better than all baseline models,and an interpretable causal graph structure can be learned on real datasets,which confirms the superiority and rationality of the algorithm.A paper has been written on the research results,and it has been submitted to the CCF A-category conference(Author order is two).Demand analysis and system design of visual analysis system.Following the requirements analysis and system design methods in software engineering,this topic analyzes the functional and non-functional requirements of the visual analysis system.Then,the overall system architecture design and function module design are completed from the whole to the part,and the database table is designed.It provides a comprehensive and reliable basis for the realization and testing of the next system.Realization and testing of visual analysis system.This subject firstly implements each functional module of the system according to the system design scheme.Then,this topic design test cases and test the system.The results show that the developed system meet the existing functional and non-functional requirements.To sum up,this topic firstly studies and implements the causal graph structure learning algorithm for dynamic networks,then conducts demand analysis and system design for the visual analysis system integrating the algorithm,and finally implements and tests the system.Finally,the complete realization of the causal diagram structure learning and visual analysis system for dynamic networks provides users with reasonable and effective causal diagram learning and convenient and intuitive visualization display functions.