This week we are broadening our scope to recapture some of the earlier concepts and expand on them. We have encountered supervised and unsupervised in week 1. Similarly we learned about PCA in week 5. In real life applications, not every problem has a designated outcome we are trying to relate other features to or predict. As a result it is incumbent upon the data scientist to have these varied tools in her/his toolbox and know when to apply them. Suppose you have been called upon to use data to drive a decision. What steps are necessary to help you decide if you have to fit an unsupervised model to help the decision making. Please be sure to describe the type of data, method and reasons.

Now suppose you decided to go with PCA. What are some of the drawback(s) that you as a data scientist will be wary of? What would you do differently to mitigate this drawback and how would you do it?