Actually, @Koleshjr, feature engineering doesn't seem to work well with this data. We are focusing more on tuning our model parameters, that's all. We did try feature engineering, but it isn't contributing much.

Thank you , May the luckiest win haha

Clearly, this is going to be lottery.

Totaly agree with you , this is definitely going to be a lottery . @Koleshjr What have helped me to reach 0.74~ is GBDT with oversampling of the minority classes using resample from sklearn.utils . And also I have used the spectral features from the satelite images .

Thank you so much @Kouassi_Jr for sharing

a hundred percent

Thank u @koleshjr

Totaly agree with you @machine_learning

Thanks you so much,

Thanks you a lot @Kouassi_Jr for sharing