3 results about "Scale-free network" patented technology

The invention provides a network node propagation influence prediction method based on multi-feature fusion and a CNN, belongs to the technical field of network analysis and deep learning, and aims to solve the problems that an existing method depends on a single topological feature, feature fusion is incomplete, and the adaptability of the CNN and network node features is poor. The method comprises the following steps: firstly, extracting a node neighborhood network with a fixed size through a BFS algorithm based on a degree value, and generating a node propagation influence label by using an SIR model; secondly, constructing a multi-channel feature matrix which comprises a microstructure feature matrix based on a neighbor node degree sum, a node attribute feature matrix based on attention mechanism dynamic aggregation attribute information, and a macrostructure feature matrix integrated with network modularity and the like to improve K-core; constructing a training set by using a BA scale-free network, designing a CNN model containing two convolutional layers, two pooling layers and a full connection layer, and training the model by using an Adam optimizer and an MSE loss function; and finally, inputting the multi-channel feature matrix of the target network into the trained model, and outputting a node propagation influence prediction value. Experiments on four real data sets of Facebook, Hep, Figeys and Hamster show that the Kendall coefficient and the propagation coverage rate of the method are both superior to those of traditional methods such as DC, BC and K-core and a single-channel RCNN method, the optimal or advanced performance is kept under different propagation intensities, and the generalization ability is high.

This invention discloses a decision-making method and system based on social dynamics evolution and hybrid reinforcement calibration. It generates a massive number of heterogeneous intelligent agents with independent cognitive architectures, drives their "perception-cognition-decision" cycle using a large language model, introduces a scale-free network-based social dynamics model to simulate the nonlinear propagation of information and opinion polarization in strong and weak relationship networks, and uses a Bayesian structural equation model with sparse anchor data to perform posterior calibration of the simulation system. This invention can perform minute-level and full-scale rehearsals for macro-strategy deployment or micro-target testing, accurately capturing emergent group behavior.

The application provides a viewpoint evolution simulation method and system based on an improved Deffuant model, and relates to the technical field of viewpoint evolution result evaluation. The application firstly constructs an initial scale-free network between individuals, and sets initial parameters and evolution running parameters of the initial scale-free network, including initial directed contact degree; then, after adding a new node to the initial scale-free network, it is judged whether the viewpoint value difference between any two nodes in the scale-free network after adding the new node is greater than an interaction threshold value, and the viewpoint value of any node in the scale-free network after adding the new node is updated by selecting a first improved Deffuant model and a second improved Deffuant model according to the size relationship between the viewpoint value difference and the interaction threshold value; until the scale-free network after adding the new node reaches a preset number of rounds, the viewpoint evolution simulation process is ended, and the scale-free network at the ending moment and the current viewpoint value of each node in the network are output. The viewpoint evolution simulation result of the application is more accurate.