Chris is a business analyst who likes to practice data modeling in her free time.
She particularly enjoys building analytical models to achieve marketing objectives. For example: clustering models for auto segmentation, propensity models for customer lifetime value predictions, and attribution models for channel evaluations.
This is a blog for Chris to practice her analytical skills and connect with like-minded people.
All in Segmentation
Collaborative filtering is a method of making automatic predictions about the preference of a consumer by collecting preferential information from various users. The underlying assumption of this approach is that if consumer A shares the same opinion as consumer B on an issue, A is more likely to share the opinion of B on a different issue than other randomly chosen consumer.
There are several ways to conduct clustering analysis. For this study, I chose to use a partitioning method that requires that input variables have meaningful magnitude or be binary categorical variables. Except for age, the demographic variables in this data are multi-category variables, so we need to perform an extra step to convert multi-category variables into binary categorical variables.
There are many metrics to used to measure the impurity of a node. Here we’ll be using the concept of entropy to measure the impurity of a node. To compute for each attribute value category level combination, we will need the equation: P(i|X=a)*Log2(P(i|X=a) combined with IFERROR to ensure that when P(i|X=a)=0 Log2(0) the undefined value will be replaced by zero. In here, we use the formula =IFERROR((D15/$F15)*LOG(D15/$F15,2),0).
