Research and Application of Improved Community Detection Algorithm Based on Wi-Fi Location Data Association Measurement Model
Bo She,Qiang Wang,Zhe Zhang,Jingru Cui,Weijie Rao,Xiaoge Zhong
DOI: https://doi.org/10.1109/iccc51575.2020.9345111
2020-01-01
Abstract:In order to make full use of the Wi-Fi Location data on university campus, we established a Structural model of neural network, and use Data Mining on students' offline social behavior through a Community Detection algorithm. Starting from the theory and technology of the Community Detection algorithm, the paper introduces the structure of neural network, the definition of community, clustering algorithm DBSCAN*-Martingale, and evaluation criteria of Community Detection algorithm; For proposing an association measurement model based on Wi-Fi Location Data, the paper also introduces the definition of similarity. And in order to obtain the similarity measurement of the edges between nodes, we used feature engineering to extract individual, global, and temporal features combined them in some ways. This paper corrects the problem of preempting nodes in the DBSCAN*-Martingale algorithm, modify the propagation characteristics of the algorithm, acquires the overlapping Community Detection algorithm DBSCAN*-Martingale pro. At the end of the paper, the algorithm was verified on artificial synthetic networks, real public data sets and Wi-Fi data sets. The improved DBSCAN*-Martingale pro algorithm performed better than the unimproved algorithm on artificial synthetic networks and real data sets and it has increased the overlap modularity from 0.613 to 0.695 compared to the unimproved algorithm.