0
$\begingroup$

I am very new to ML and currently, I am working on building a model that can predict recurring blood donors (a classification problem). I have a dataset which consists of 25 features (gender, height, age, previous donations, etc).

However, this data is not labelled. But, I was thinking of considering the ratio between the previous donation count of a donor and their age to label my data, and using a threshold value to classify whether the donor will come back to donate blood or not. For example, given a donor is 25 yrs old and has donated blood 20 times. So, the number of previous donations divided by the age of the donor equals to 0.8. So, if the threshold value is 0.55, then I would label this instance as a 1 (this is a recurring donor).

So, can I label my data using this technique? Or else, should I use some unsupervised learning model (like clustering)?

I have selected the important features from my dataset and I have cleaned up the data. I want to now train a model, but I am a bit undecided as to whether I should use unsupervised learning techniques or not. This is because it is sometimes difficult to derive meaning from the results of unsupervised learning models.

$\endgroup$

1 Answer 1

1
$\begingroup$

Whether you can do this is going to depend on what "recurring donor" is supposed to mean. Defining it based on age and number of donations seems pretty sensible to me, but, in the end, it's going to depend on what you're trying to do with the model (i.e., what you intend for "recurring donor" to mean).

This seems like a situation where you should consult domain experts, if you aren't one yourself: you could look for more formal definitions given by organizations for what a "recurring donor" means.

Once you have this, you can then consider whether this method of synthetically generating labels aligns with that definition. This seems like a textbook classification problem to me, so coming up with a set of labels would be ideal.

Even if you do end up using unsupervised techniques, to make sense of the results (i.e., for finding distinguishing features in recurring donors), you'll still need a solid definition of what a recurring donor is, so this is what you should start with.

$\endgroup$
1
  • 1
    $\begingroup$ Thank you! I will try consulting a few profs before moving forward with this project. $\endgroup$
    – stkmnd
    Sep 4, 2023 at 23:10

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .