A Data Mining System for the Generation of Geographical C16 Cancer Patterns

This paper describes the results of data mining system developed ad hoc to address the problem of discovering patterns of interest in population databases for cancer. In particular, the experimental results obtained by the system are shown. The architecture of the system is innovative since it integrates a visual cartographic, a data warehouse and a data mining system. The k-means algorithm was used for the generation of patterns, which permits expressing patterns as regions or groups of districts with affinity in their localization and mortality rate parameters. The source databases used in this investigation were obtained from Mexican official institutions. As a result of the application of the system, a set of clustering patterns was generated, which defines the mortality distribution for stomach cancer in Mexican districts.