is minimal. Other scoring functions are possible and more general
regression models or even neural networks can be used. However, these techniques
are out of the scope of this book and the interested reader is referred to [52].
Classification requires a categorical response variable. In some cases it makes sense to transform a numerical response variable into a categorical one. For example, for Table 3.1 one could decide to transform variable age into a categorical response variable by mapping values below 70 onto label “young” and values of 70 and above onto label “old”. Now a decision tree can be constructed to classify instances into people that die(d) “young” and people that die(d) “old”. Similarly, all values in Table 3.3 can be made categorical. For example, positive values are mapped onto “true” (the item was purchased) and value 0 is mapped onto “false” (the item was not purchased). After applying this mapping to Table 3.3, we can apply classification to the coffee shop data while using e.g., column muffin as a response variable. We could, for instance, find that customers who drink lots of tea tend to eat muffins.