Fundamental concepts: Our fundamental concepts as the basis of many common data science techniques; The importance of familiarity with the building blocks of data science.
Exemplary techniques: Association and co-occurrences; Behavior profiling; Link prediction; Data reduction; Latent information mining; Movie recommendation; Bias-variance decomposition of error; Ensembles of models; Causal reasoning from data.
As discussed in the previous chapter, a useful way to think of a team approaching a business problem data analytically is that they are faced with an engineering problem—not mechanical engineering or even software engineering, but analytical engineering. The business problem itself provides the goal as well as constraints on its solution. The data and domain knowledge provide raw materials. And data science provides frameworks for decomposing the problem into subproblems, as well as tools and techniques for solving them. We have discussed some of the most valuable conceptual frameworks and some of the most common building blocks for solutions. However, data science is a vast field, with entire degree programs devoted to it, so we cannot hope to be exhaustive in a book like this. Fortunately, the fundamental principles we have discussed undergird most of data science.
As with other engineering problems, it is often more efficient to cast a new problem into a set of problems for which we already have good tools, rather than trying ...