Customer Propensity Analysis - Scheduling Service - Variavel_target

Asked

Viewed 54 times

-4

I’m starting to take the first steps in machine-learning and would like to apply a prospecting analysis to each client according to their scheduling service.

I have a client portfolio that has one or more active products. Each period - approximately 6 months - the portfolio needs to be maintained.

The company has a proactive service for scheduling the service. The customer can choose to schedule the service or go directly to technical assistance.

The objective of the study is identificar which clients have greater prospect for such scheduling.

Example:

Consider a customer X with 2 products - Prod_a and Prod_b - and which required 4 maintenance in the year - Prod_a-Man_01, Prod_a-Man_02, Prod_b-Man_01 and Prod_b-Man_02. However, I have doubts in the preparation of the data. The doubts are as follows:

  1. What should be my variable target?

The target variable "Scheduled: Yes/No" for each customer maintenance or should consider the client only 1 time and work with the target variable "% Scheduling" counting how many times the client scheduled the total of 4 maintenance that he had to perform.

  1. What models are indicated for this type of analysis?
python machine-learning

  • It is always important to have the business problem well defined before you start. There is no 'right and wrong', there is what you want as the end result. Without knowing your data it is difficult to propose something because I do not know the other variables and do not know which ones are important, without your data set it is also difficult to tell which machine model Learning will adapt best to the proposed problem. If your business rule is X you can group, if it is Y you can consider the customer only once. It all depends on your business problem and business rule. Hug!

1 answer

0

To get a better idea, just conducting an exploratory analysis. I have in mind some questions, such as:

  • The profile of the client who made the scheduling is related to products individually or generally?
  • Which (or which) item has the most relevance with client scheduling?
  • Which (or which) item generates less engagement?
  • Gender, age, etc., influence engagement?

From this exploratory study and plots, you could understand what behavior of each customer for each product and, only from there, find a suitable model. The most basic would be a logistic regression or even a decision tree, applying some treatment under the variables.

Especially in the decision tree, you might be able to simplify by transforming into qualitative variables. If it helps, take a look at a basic example in my github

Browser other questions tagged python machine-learning

You are not signed in. Login or sign up in order to post.