Identify earthquake hot spots with 3-dimensional density-based clustering analysis

Clustering analysis to identify hot spots of interest is a popular topic in spatial data mining. One of the challenges for existing clustering approaches is that they are not suitable for clustering high-dimensional feature vectors. Many existing density-based clustering algorithms model the point density based on two-dimensional information without considering the impact of the vertical dimension. This paper proposes a revision to the Supervised Clustering Density-based Estimation (SCDE) algorithm by introducing the depth variable into the influence function. The paper then develops an experimental design exploring the impact of depth on SCDE, and measures the reward and fitness functions as the evaluation criteria. The paper also develops a metric to assess the accuracy of estimating high-risk earthquake spots by overlaying the clustering results with the California Seismic Hazard Map. Results show that treatment with the depth variable generates better fitness values, rewards, and accuracy.