feat(regressor): add regressor interface

[tczhao]

add regressor interface

resolves #14 #4

[xuyxu]

Brilliant job! Thanks @tczhao! I will take a careful look tomorrow ;-)

[tczhao]

The classification performance on our highly skewed data is not that good but the regression performance is incredible

[xuyxu]

Hi @tczhao, I have left some comments on your PR, let me know if you are OK with them. Thanks for this great PR, I will work on the experiment parts on the regressor and let you stay stuned. Hopefully, this PR can be merged in a few days.

[xuyxu]

The classification performance on our highly skewed data is not that good but the regression performance is incredible

Thanks for sharing this. Another question is that what is the final number of cascade layers in your regression dataset? If the number of 1, I am afraid that the cascade structure does not have any positive effect, and the entire model is simply an average of forest models...

EDIT: I have added the experiment on regression, you can check it via the last GitHub checks

[tczhao]

The classification performance on our highly skewed data is not that good but the regression performance is incredible

Thanks for sharing this. Another question is that what is the final number of cascade layers in your regression dataset? If the number of 1, I am afraid that the cascade structure does not have any positive effect, and the entire model is simply an average of forest models...

EDIT: I have added the experiment on regression, you can check it via the last GitHub checks

Thanks for the added tests

It's 3 layer in the end (not just a simple ensemble), we set the max to 3 for some quick experiment

[xuyxu]

LGTM

[xuyxu]

Thanks for your great work @tczhao! I will merge the PR soon ;-)

[xuyxu]

Merged.