bestkao / customer_segments Goto Github PK

First of all, thank you for making this awesome notebook public, I work on a similar dataset and I am so grateful to be able to look into your work!

Concerning my issue: I think I found an inconsistency in your discription of the principal components' feature weights. In the answer for question five you describe for example the weights on the first PC:

In Dimension 1 a significant positive weight is placed on Detergents_Paper with meaningful positive weight on Milk and Grocery. This dimension is best categorized by customer spending on retail goods.

The table row for dimension one looks as follows:

Explained Variance | Fresh | Milk | Grocery | Frozen | Detergents_Paper | Delicatessen
0.4430 | 0.1675 | -0.4014 | -0.4381 | 0.1782 | -0.7514 | -0.1499

You seem to attribute a positive weight to a dimension's feature if the feature weight is negative and a negative weight if the feature weight is positive. Is this correct? It seems very counter intuitive and contradicts the hint you give:

A positive increase in a specific dimension corresponds with an increase of the positive-weighted features and a decrease of the negative-weighted features. The rate of increase or decrease is based on the indivdual feature weights.

Again many thanks for your great work.
Sincerely,
Mattzee