Gender-Based Violence Tweet Classification Challenge
Helping Global
2000 Points
Challenge completed almost 4 years ago
Natural Language Processing
Classification
634 joined
140 active
Start
Aug 09, 21
Close
Nov 14, 21
Reveal
Nov 14, 21
How professor got this result?
Help · 14 Nov 2021, 01:13 · edited 2 minutes later · 11
How to got this result ?
Impossible from this data I used all approach and techniques in ml and DL the best result got 78%
There's still much time to beat the result, the key is balancing the labels via any popular approach, for me, I used pseudo labels.
I used many approaches but need so much from time to balance data in the deep learning approach
Hmm, I see, keep it coming.
saw your solution but is it permissible to use test data in the training process
Yeah sure, although you can only use datasets provided, excluding meta data, as well as open source pretrained models.
But you using test data in training
Lol, you are open to use all provided data as you deem fit.
But do not use the test data in the training process. This is considered cheating
Focus more on learning @qusaibenferas, that's why I made the solutions available. Pseudo labeling is a technique widely used in data science competitions, it's in no way considered as cheating, it's called semi supervised learning.
Feel free to reach me or check out the learn tab on Zindi to find insightful ways of building machine learning models.
I mean, test data is allowed to be used in the training process
It's not against the rules of the competition. Keep winning @qusaibenferas