I have a question about User segmentation I have a data-set of rides of an application that works at the same model as Uber. The attributes I have : Reservation it means the id of the Ride and statutCourse means the status of the ride cancelled or finished ..... the Id client and the entreprise is the company where the client works IDChaufeur is conductor ID and i have the geographic coordinations of the start point and the finish point i also have date and hour of the ride also real time and estimated time and estimated distance

enter image description here

I want to classify/cluster clients profiles and conductors profiles

My problem is that one client can make many rides and I don't know if I can use many lines of rides for the same client in one dataset and than use this dataset to classify the clients into classes or profiles ?


1 Answer 1


If you want to cluster people, you need to have one row per person. To do this you will need to group by IDClient/IDChauffeur and compute features like meanHourRide, stdHourRide, ...

By feeding the algorithm with multiple records of the same person (with the same travels for example), the clustering algorithm might create a cluster with similar travels (time/location/distance-wise) even though it is made by only one customer.

This would result in travel clustering, which is also possible but not what you want here.

  • $\begingroup$ Thank you very much for your answer, I think i should have a **one row per person ** but if i want to use geographic coordinates that belong to one person how could I do that since one person has many rows(rides)? like find out form where and to where this client mostly takes rides can you help find a way to do that? $\endgroup$ Commented Apr 11, 2020 at 11:00
  • $\begingroup$ You could compute features like : - the distance of pickup point to the city center - the distance of drop point to the city center - difference between those two Then you can compute the mean/std of those variables for each client/driver. If the pickup points and drop points are grouped into 'interest zone' commercial zone, theatre, etc. you could create features "theater", "city center", "mall", which take the value of how many times the user has been taken or dropped there. If my answers helped, please click the "accept" button on the left. $\endgroup$
    – Rusoiba
    Commented Apr 11, 2020 at 14:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.