Classification techniques have shown recently their usefulness for complex process as high dimensionality diagnosis. Also the detail that no physical model for the process is mandatory, they enable to study the problem of sensor location. Preliminary made previously in the area of living process diagnosis have been the initial key point to extend its application to the medical diagnosis framework. In spite of the behavioural difference, both domains display many common practices. However, medical diagnosis recently has carried grave challenges such (gene expression profiling) and heterogeneity of data (symbolic histo-pathological factors). We show here that both challenges can be overawed and used in return to improve complex process diagnosis.

KEYWORDS: Data mining, Naïve Bayes Classifier, k-Nearest Neighbor Classifier.