Spatio-Temporal Data Mining for Hotspot Identification and Evolution in Health Informatics

Abstract

Health emergencies in the recent years have alarmed every country to do preparedness before any critical situation. The pandemics in recent years have disrupted nearly every human. Using hotspot identification techniques timely measures can be made at affected areas. The previous papers were mostly related to some specific diseases hotspot areas identification. No work was done on multiple diseases hotspots and their close proximity with each other. The focus of this study is to discover multipe disease events occurring in close proximity in space and time and also identifying the trajectories of single disease in different times. The available dataset conatains 3 major diseases data, TB, Malaria and AVH. All these are major occuring diseases in punjab province. Dataset is containing data of 36 districts of punjab province. In this paper K-Means clustering technique is used to make clusters separately within every district. Each district have 1-4 clusters representing tehsil level cases and district level cases. After making clusters in each district, connections were made between the clusters of these diseases to check the close proximity between diseases on same time interval. Proximity is also checked for four years time interval making 48 intervals separately for each month. Analysis is made on the different trajectories of single diseases. Single disease hotspots are also checked if they are making any seasonal or temporal trend or any spatial based trend. If a single disease cluster is either making a persistent hotspot or a new hotspot or a diminishing hotspot based on the condition that how this cluster is occurring in past and most recent months. In results, the close proximity is identified between different diseases clusters and also categorize there connections. A connection between any of two diseases clusters is either high-high, low-low, high-low or low-high. These connections were based on the number of patients in a cluster and the distance of cluster from the other disease cluster. Also, the trajectories of single diseases are identified and the clusters that are persistent.