Clustering Data Cuaca Ekstrim Indonesia dengan K-Means dan Entropi

Abstract

As information technology develops in agriculture, more patterns of database systems are manual or computerized. However, the amount of data available does not always match the knowledge that can be generated. The community really needs weather information regardless of the format if it is reliable and valid. One is to determine local climate patterns. The data collection method uses a document study method, then uses data analysis techniques with a scoring system on rainfall data, maximum temperature, minimum temperature, average temperature, humidity, and wind speed. The K-Means Clustering method is a technique for grouping data. From the analysis carried out, there are 3 classes to cluster the weather levels produced by the entropy test to avoid bias towards non-optimal precision and accuracy. The division of the number of clusters can be captured as a type of potential vulnerability, namely high, medium, and mild where the total of all weather data from all BMKG stations throughout Indonesia is 123 data, 52 of which are classified as areas that have the potential to experience extreme high weather, 31 are classified as areas that experience extreme weather, moderate extreme weather potential, and 40 areas classified as areas with mild extreme weather potential.