good luck, Did you try the transformer model or tree?

I haven't tried transformer models

same here.

hi krishna welecome back to zindi, i think feat eng working for all of you but here is different may be i have to find some secret.

@yisakberhanu Even I am not able to find any magic feature as of now :). But tree models definitely did not work for me. statistical time series models are performing better for me. I have not explored neural networks as of now.

Yes, but the data is not quite good. The main thing here is predict in future time but the provided data is not right way.

I don't know about the others but for Me I feell like the data is quite good

Agree, this dataset is rich with patterns that can be helpful to extrapolate the future (trend, seasonalities). In my opinion it's all about how you extract those patterns with thoughtful feature engineering

that great insight, but , is tree model working for you ?

always itu competition has problem on data structure, you know the use cases is not same like this, this is for only research and this not applicable on real world.

but this one doesn't have a problem , trust me. Even the Cv is very very reliable!

@Koleshjr agree again, this dataset is well structured and CV is reliable

@Yisakberhanu based on my experience it's less about model type, more about data massage

wow that is great but for me feat eng is not working, what do you think about transformer?, did you use it ?