Code Solution for machine learning problems solved in AI Hackathon of Tunisia.
in order to use, execute the code in the repository youi will be needing python enviroment, I prefer using anaconda with XGBoost and LightGBM models.
In this repository you will find the code solution for both Individual and Startup Track.
In the TUnis air problem we used the LightGBM model without grid search, and in the Startup we used CNN model which called DenseNet, pretrained on the ImageNet Dataset
Doing a Machine learning Hackathon within 20 hours is just a nightmare, I learned that you need to preprocess the data and do basic feature engineering within the first hours, then fine-tuning of the final algorithm and more feature engineering will be the key aspect for your solution to win.