Taxonomy

Abstract

Feature selection, which was discussed in the last chapter, is a powerful component of the DOPA process. It enables the tuning analyst to quickly identify areas of the database that are performing outside of normal. The metrics with a high incidence of flagged values are assumed to have a high predictive value of pointing to the problem area. And this is definitely true in my experience. While the feature selection/flagging process is sufficient by itself to solve many problems, I learned another analytics “trick” from my son that enabled me to take my analysis one step further. The concept I brought into the analysis is that of taxonomy.