How we built it
For first task we used several models to verify and found out that xgb regressor is giving the least mean squared error , We found out that the most important feature is Smoking History which have the correlation of 0.97.
In the second task we initially developed a custom CNN architecture, achieving an accuracy of 59.8%. Subsequently, we fine-tuned a pre-trained model, EfficientNetV2, boosting our accuracy to 75%. Through additional techniques such as image sharpening and augmentation, our accuracy soared to 82%.
Built With
- python
- tensorflow
Log in or sign up for Devpost to join the conversation.