Distribution Model of Personal Protective Equipment (PPE) Using the Spatial Dominance Test and Decision Tree Algorithm

The COVID-19 case has developed positively, but preventive measures must be taken to anticipate SARS-CoV-2 mutations. Anticipation can include policies, preparing health workers, and providing personal protective equipment. Personal Protective Equipment (PPE) availability is a big challenge in handling pandemics, especially COVID-19. The level of need for PPE in an area depends on the number of COVID-19 cases. This research provides a solution to overcome the availability of PPE by applying the concept of cross-regional collaboration. Areas with low COVID-19 case rates can help areas with high COVID-19 case rates by sending PPE assistance. Implementing the cross-regional collaboration concept is assisted by the spatial dominance test algorithm, namely the spatial skyline query. Spatial Skyline Query works by searching for the most ideal area. The ideal area is an area with low COVID-19 case criteria. The low number of positive cases, death cases, probable cases, and close contact cases supports the low number of COVID-19 cases. Areas with the highest number of recovered cases are also priorities. The SSQ model was developed into two models for searching priority areas for PPE assistants. The first model is Sort Filter Skyline 1 (SFS1), and the second is Sort Filter Skyline 2 (SFS2). SFS1 is a form of SFS algorithm optimization that searches for the best 50% of all regions. SFS2 modifies SFS1 by selecting areas whose distance is <= the average distance of the area to the Health Crisis Centre of the Ministry of Health of the Republic of Indonesia. This research involves searching for priority areas and applying a prediction algorithm to extract knowledge built from the prediction model. The algorithm used is C5.0. The data used to apply the prediction algorithm results from the application of SFS1 and SFS2. The results of testing the prediction model by the C5.0 algorithm produced an accuracy of 77.26% for SFS1 data and 92.01% for SFS2. The average rules resulting from the C5.0 algorithm are three for SFS1 and two for SFS2.