Last answered:

17 Mar 2021

Posted on:

08 Mar 2021

0

Can I perform the purchase analysis with real life data that doesn’t contain demographics?

Hi, I’m trying to conduct similar analysis to the purchase data performance but I do not have customer demographic data attached to the purchase data that I have to hand. Can I still conduct meaningful analysis on it? Perhaps you can point me to a video that allows me to deep dive on the purchase data analytics (multiple products, only tracking transactions and not tracking when they browse the website) to come up with similar segmentation?

3 answers ( 0 marked as helpful)
Instructor
Posted on:

11 Mar 2021

0

Hi Michael,
even you don't have demographic data you should still be able to perform data segmentation. However, you will not be able to use the exact same fitted kmeans pca algorithm, as it segments the data based on the demographic features in data.
What you might want to do instead is segment the data you have at hand. You could split your available data and split it into train and test data and perform segmentation. Afterwards you'll be able to analyze the clusters and perform purchase analytics.
Keep in mind that the quality of the data and the features is also key - there is data which is even better suited for segmentation compared to demographic data. But there is some data which might not be revealing when it comes to different types of consumers. So, that is something to factor in, when choosing the features for your segmentation model.
Hope this helps!

Best,
365 Eli

Posted on:

12 Mar 2021

0

Thanks Elitsa!
Is there a video in this site that mirrors the methodology you are promoting?

Instructor
Posted on:

17 Mar 2021

0

Hi Michael,
The methodology would be the same, even if the data changes. What's crucial is to understand the data and choose the important predictors for your problem. That's also what I consider one of the main challenges in machine learning, as it is often difficult to determine these predictors just by looking at the data. But it does get easier the more data sets and problems you tackle, as you develop a better understanding of how to progress.
As far as content from our program goes:
You can take a look at another PCA example from the machine learning course, as it is also about customer segmentation:
Machine Learning in Python | 365 DataScience
(The previous chapter is also about cluster analysis)
We will be featuring PCA in one of our upcoming courses, however it will not relate to customer segmentation or purchase analytics. Instead, it will be a part of a Linear Algebra and Feature Selection course, where we'll compare and discuss approaches for dimensionality reduction and feature selection. We'll feature data sets from different industries.

Best,
365 Eli

Submit an answer