n*********e
发帖数: 318 | 1 ------
Say, I have 10,000 customers and their data from a merchant.
Currently, I have successfully built two predictive models to predict for
each customer -
1) predicted prob. of a purchase (buy vs. no-buy)
2) expected spend $ of this customer given it is a buy
--------
In the end, I need to deliver just one score for my merchant client so that
they can approach their customers starting from top of the list (and working
down the list).
How should I construct a single score to combine both information?
Thanks! | k****n
发帖数: 165 | 2 I'm wondering how you handle the cutoff point for your model (1).
Essentially every customer has a 2-dimensional vector. The first entry is
his prob to buy while the second one average $ spent.
One possible way to create a scaler statistic is to 1) calculate the
percentile of the average $ spent. 2) report p * w * q, where p is his prob
to buy, q is the quantile of his expect $ spent, w is your relative weight
given to miss a potential whale customer. | l******n
发帖数: 9344 | 3 no-buy client has 0 expected expenditure. If you have purchase quantity and
unit price, you can just predict the total expenditure as the product of
quantity and unit price.
that
working
【在 n*********e 的大作中提到】
: ------
: Say, I have 10,000 customers and their data from a merchant.
: Currently, I have successfully built two predictive models to predict for
: each customer -
: 1) predicted prob. of a purchase (buy vs. no-buy)
: 2) expected spend $ of this customer given it is a buy
: --------
: In the end, I need to deliver just one score for my merchant client so that
: they can approach their customers starting from top of the list (and working
: down the list).
|
|
