Exam AWS Certified Machine Learning - Specialty topic 1 question 50 discussion

All of the questions in the preceding examples rely on having example data that includes answers. There are times that you don't need, or can't get, example data with answers. This is true for problems whose answers identify groups. For example: "I want to group current and prospective customers into 10 groups based on their attributes. How should I group them? " You might choose to send the mailing to customers in the group that has the highest percentage of current customers. That is, prospective customers that most resemble current customers based on the same set of attributes. For this type of question, Amazon SageMaker provides the K-Means Algorithm. https://docs.aws.amazon.com/sagemaker/latest/dg/algos.html Clustering algorithms are unsupervised. In unsupervised learning, labels that might be associated with the objects in the training dataset aren't used. https://docs.aws.amazon.com/sagemaker/latest/dg/algo-kmeans-tech-notes.html THE ANSWER COULD BE B.clustering on customer profile data to understand key characteristic

Option C. This is not purely unsupervised, as clustering would be, because we have current and past customer profiles to go on. We want to find new customers by finding similar profiles on social media. So it is supervised to some extent. It's not a cluster problem; it is user-user collaborative filtering. The key is to recognize that this is not clustering. You're not blindly trying to group people. You have existing profiles that you are comparing them to.

'B' is correct

It is B. Recommendation Engines: Traditionally focus on suggesting products/services to existing customers based on past behavior.

Clustering is right

C would be an answer if wanted to send the promo to the existing customers. But we want to find potential customers. And we can do it only by comparing existing customers with potential customers. It can be done by creating clusters of existing customers and measuring the distance to those clusters for the new potential users. So my answer is B

A. NO - Linear Regression not best to understand relationships between data B. NO - it is supervised (we know premiums received vs. claims paid, so can assign users to GOOD or BAD), so no clustering C. YES - A recommendation engine in AWS lingua is Amazing Recommender (https://docs.aws.amazon.com/personalize/latest/dg/what-is-personalize.html - "Creating a targeted marketing campaign") and can create user segments D. NO - not as good as C

B for me

Recommendation engines is perfect for customers we have, but for implementing a machine learning model to identify potential (new customers on social media) this requires clustering and segmentation. https://neptune.ai/blog/customer-segmentation-using-machine-learning

Based on the link below, it must be C https://medium.com/voice-tech-podcast/a-simple-way-to-explain-the-recommendation-engine-in-ai-d1a609f59d97

We are divided, but I stick with B.

I think it should be c

recommender system would help here, as we already have details of all customers

it should be C - recommender system would be better fit here.

We should use recommendation system to find key characteristics only among company users (past and present). At this step we don't take any users from the web. After we finish processing this CF model we identify key characteristics (important features?) and only after that, we will start looking for similar users on the web.

I would use clustering technique to identify which customers in my database are the target audience and get similar customer profiles from the social media dataset. Its a lot simpler

recommendation engines can use either supervised or unsupervised learning. I can't find any reason to NOT use recommendation engine???