I have analyizing the data for students prediction using their marks.
Whether student is bad or good
I am not getting satisfying result.
Mey anyone help me what to do?
Hello Nirav,
Thanks for sharing. I see a couple of issues based on the screenshot shared:
- There are 6 claases but only 15 total samples. If this is training data, this is very little data for a multi-class classification problem.
- For “less strong students”, it shows highest metrics, meaning the model has only learnt(or have memorized) this class.
How to resolve:
- Please be clear of the end goal first. If the question is to solve “Whether student is bad or good”, then it is a binary classification problem. Classifying students into 6 categories as shown here would be amulti-class problem.
- Also the more the training data, the better the model learns. If not much data available, try synthetic data generation methods like SMOTE.
- Then, make sure to have balanced data(may be around 50% of each class) to make sure the model sees good amount of data for pattern recognition in each class.
- Make sure normalize features.
- Though the model used is invisible here, it is a best practice to start with the simplest models first(like logistic regression if binary).
Hope this helps!
1 Like
Could I share the data or csv file here?
It is a multi-class problem.
I use scalar then logistic regression it doesn’t give the result I look for.
I paste the code here.
Mey you please go through my code?
`[type or paste code here](https://drive.google.com/file/d/1HtC7Imtgscrr2LV27jUx75PyRbX3wmMS/view?usp=drive_link)`