Research On Density-Based Clustering Algorithms And Its Application In Inclusive Finance

Density-Based Spatial Clustering of Applications with Noise(DBSCAN)is a typical density-based clustering technique that can detect arbitrary shaped clusters and has been utilized in many applications because it does not require the number of clusters to be specified and is robust to noise.It is frequently used in picture segmentation,pattern recognition,market research,financial analysis,and a variety of other applications.The DBSCAN clustering method is explored in this study,its reliance on parameters and other issues are addressed,and it is applied to financial analysis,focusing on the following research:(1)To address the problem that the DBSCAN algorithm relies on the global parameters Radius Threshold eps and Density Threshold Minpts,this paper proposes a new Density clustering algorithm for Adaptive Radius and Density Thresholds(ARDT-DBSCAN).The algorithm firstly obtains the close neighbours based on the intersection of the k-nearest neighbours and the reverse k-neighbours of the data points;secondly,it obtains the mean value of the close neighbours,i.e.the adaptive density threshold,based on the mean value of the close neighbours,then screens the core points based on the mean value of the close neighbours,uses the average distance between the core points and the close neighbours to obtain the adaptive radius,and clusters the core points;finally,it uses label transfer to assign the clustering labels of the remaining points.The experimental results show that the ARDT-DBSCAN algorithm not only achieves adaptivity,but also outperforms the DBSCAN and ICKDC algorithms on both artificial and real datasets.(2)To address the problem that the DBSCAN algorithm adopts a “first-come,first-served” strategy for edge clustering,which affects the stability of the algorithm.In this paper,a Density clustering with Adaptive Multi-step Assignment mechanism(DAMA-DBSCAN)algorithm is proposed on the basis of ARDT-DBSCAN.The algorithm first uses the ARDT-DBSCAN algorithm to obtain the core points and clusters the core points to obtain the initial clusters,further identifies the remaining points as edge points and noise points,and uses the inverse of the close neighbourhood mean and connectability to assign the edge points;secondly,it uses a re-checking mechanism to deal with the edge points in the overlapping region;finally,the final cluster labels are assigned to the noise points.The experimental results show that DAMA-DBSCAN outperforms DBSCAN and ICKDC algorithms on both artificial and real datasets.(3)In order to address the problem that the huge amount of data on low-income customers of banks in urban and rural areas,the large number of geographical intersections and the complexity of overlapping areas can seriously affect the ability of clustering algorithms to segment low-income customers of banks,this paper uses the proposed algorithm DAMA-DBSCAN to carry out clustering analysis on the basis of the principle of low-income bank customer segmentation,and develops the “Inclusive Finance Knowing You” segmentation assistant to analyse the results,which provides a new idea and method for the promotion of inclusive finance of banks in urban and rural areas.