The Curious

""" How to Handle an Imbalanced Dataset? """

What questions should one ask about the Imbalanced Dataset before proceeding?

No, Oversampling isn't the only answer, at least not in every scenario.
As a research paper mentioned, "using oversampling techniques to artificially balance the dataset—by replicating the minority class or generating synthetic samples—can be counterproductive."

Here are a few pointers I found on the Internet

1. Recall or Precision, decide what is important for you.

2. Try to start with under-sampling the majority class

3. Feature engineering is important

4. Try Focal Loss instead of Log Loss

5. An ensemble of XgBoost and LightGBM may work best

6. Don’t upsample/downsample. Train using log loss, then just tune the decision threshold based on the respective costs of false positives and negatives

Getting started on Computer Science without Money and Certification.
Guide yourself through like the Doggo:

1. MIT OPEN MIT Open Course Ware

https://ocw.mit.edu/courses/6-0001-introduction-to-computer-science-and-programming-in-python-fall-2016/

2. OPEN SOURCE SOCIETY UNIVERSITY - https://ocw.mit.edu/courses/6-0001-introduction-to-computer-science-and-programming-in-python-fall-2016/

3. Stanford Machine Learning by Andrew Ng - https://online.stanford.edu/courses/cs229-machine-learning
[Available on Youtube as well]

4. Project-based free learning - https://www.freecodecamp.org/

5. Harvard CS50 - https://pll.harvard.edu/course/cs50-introduction-computer-science [Available Online]

6. Web Security Academy by Portswigger [Creators of Burpsuite] - https://portswigger.net/web-security/learning-paths

7. Khan Academy - https://www.khanacademy.org/ [Not just for CS- for everything]https://www.khanacademy.org/

What's wrong with gemini, is it drunk?

Specialization is for insects.

Robert A. Heinlein
Time Enough for Love (1973)

[Image collected from Princeton.edu]

Limbo (2020)
Director: Ben Sharrock