Look Alike Modelling


#1

Could anyone please let me know how to do look alike modelling?
My Business question is to find out the prospective customers for targeting. These customers have not mad any transaction with us so far(they are new customers). I would like to build a look alike model using the existing customer data and target new customers.

Please let me know how to go about this!!


#2

@vijay1987,

You can deal with this challenge in following ways:

  1. You have input and output variable for existing customer (train data set) and only input variable for future customer (test data set). Now, you will build a model based on train data set and apply this
    model on test data set to predict the behavior
  2. Build different segment of customer based on input features only (known as segmentation) and identify the segments/ clusters of customers using unsupervised modeling techniques (K-means…). Hypothesis behind this technique is, all the customers of same clusters will behave similar compare to other clusters.
  3. You can combine method 1 and 2 together for better prediction power also. Build clusters and build different model for each clusters and predict the behavior of customers.

Hope this helps!

Regards,
Imran