1. Data Mining

Typhoon data mining we deal with in this paper is a data mining for a scientific domain on a meteorological application in the form of image / multimedia data with spatio-temporal properties. Hence many types of data mining algorithms can be applied to this data collection, and we may need an extensive and comprehensive study to determine which algorithms work best for this particular application. To review various data mining algorithms, we classify them into three categories.

  1. Spatial data mining
  2. Temporal data mining
  3. Spatio-temporal data mining

Spatial data mining deals with the two-dimensional distribution of typhoon cloud patterns, but note here that a feature space for two dimensional spatial patterns has, in general, much higher dimensions than two. Temporal data mining, on the other hand, focuses on the temporal dynamics of typhoon cloud patterns and involves the modeling of the life cycle of the typhoon. Relatively speaking, spatial data mining is more concerned with typhoon analysis, while temporal data mining is more concerned with typhoon prediction. Spatio-temporal data mining integrates both types of data mining, and therefore should be most powerful, but we are yet to develop or test algorithms of this category because of the complexity of the data collection and algorithms.