Research On Overlapping Community Detection Algorithm Based On Adaptive Truncation Distance

The constituent relationships of many things in the real world can be abstracted into a complex network model,of which community structure is an important characteristic in complex networks.Community structure is a collection of individuals with common characteristics and is widely present in complex networks.Probing the community structure in complex networks helps us to understand the formation and evolution mechanism of the system,and can make predictions accordingly.Different communities often interact with each other through overlapping nodes,and communities are not completely independent of each other,so it is more relevant to study the structure of overlapping communities.The Density Peaks Clustering(DPC)algorithm is more efficient than other traditional community detection algorithms to obtain the community structure by clustering,and this algorithm has received much attention.However,DPC needs to include the distance information between all nodes as input and cannot directly deal with complex networks represented by adjacency matrix;DPC introduces truncation distance when calculating the local density of nodes,which is usually set as a fixed value based on experience and lacks adaptiveness for different network structures;DPC requires manual participation in selecting clustering centers and cannot select clustering centers adaptively;in addition,DPC DPC does not consider overlapping nodes in the assignment strategy,so it cannot detect overlapping nodes and cannot meet the demand of overlapping community detection.In order to exploit the advantages of DPC in terms of speed and efficiency and to compensate for the current shortcomings of DPC.In this paper,we extend DPC and propose Extended Adaptive for Detection of Overlapping Community(ADPC),a new distance function that converts the adjacency matrix into a distance matrix containing distance information between all nodes that DPC can handle.ADPC proposes a new distance function that converts the adjacency matrix into a distance matrix containing distance information between all nodes that can be handled by DPC,and the distance function can be generalized to networks with or without weights.ADPC proposes an algorithm that adaptively computes the truncation distance based on the structural properties of the network.In addition,unlike DPC,which requires manual participation in selecting the clustering centers,ADPC uses the k-means dichotomous clustering idea to iteratively process the adaptive selection of clustering centers without manual participation.To support DPC to detect overlapping nodes,ADPC proposes an algorithm to detect all candidate overlapping nodes in the network and adds overlapping node assignment on top of the DPC assignment strategy,making ADPC support the detection of overlapping nodes.The effectiveness of adaptive truncation distance for ADPC was demonstrated by setting the truncation distance to a constant value and adaptive value respectively in the experiments;the effectiveness of ADPC in discovering overlapping communities was verified by running on real datasets and LFR synthetic networks;the comparison with comparative algorithms such as CPM,LFM,SLPA,DEMON and MOSES on LFR synthetic networks showed that ADPC performs better on networks with complex network structures or complex weight distributions;the effectiveness of the candidate overlap node detection algorithm for overlap node detection in complex networks and the impact of the parameters introduced in the overlap node assignment strategy on overlap node detection are verified by designing experiments on the parameters of overlap nodes on LFR synthetic networks.