The Rank-Based Detection Algorithm (RBDA) is an outlier detection method using unsupervised learning. If we use the RBDA, a certain hyperparameter must be specified with the advanced knowledge that the detection accuracy will change depending on it. However, it is difficult to specify this value from the data in unsupervised learning. Therefore, in this study, we proposed to adopt the ensemble method to address this problem. We compared and examined the accuracy of the proposed method in various situations through simulations and actual data analysis. The simulation results show that the accuracy of the proposed method is higher than that of ordinary RBDA in many cases. It is also found that the accuracy of the proposed method is not significantly impaired in all situations.
In recent years, analysis using spatial data (data with geographical location information) has been carried out in many fields, and research is being actively conducted in the field of spatial epidemiology. In this study, we used R shiny to develop software that can calculate risk indicators that represent the risk of illness and death, and visualize those results on a map. This made it possible to spatially grasp the distribution of risk indicators while performing interactive operations. In addition, this software also implements a function to hotspot detection in order to evaluate whether or not high risk areas are concentrated in a specific area. In this paper, we describe the risk index of death and hotspot detection, while introducing the software we developed to perform comprehensive analysis.