I fit a many different models to my data set and I still get horrible accuracy. Any suggestions?

Question

I'm trying to create a regression with a data set that does not contain any nulls and has a very few outliers. I fit a linear regression, a random forest, and a gbm model but they all have terrible accuracy.

Any suggestions on how to move on from this point? Feeling like I've hit a road block.

Thanks!

Answer 1

Make sure you have done preprocessing properly, features are standardized and normalized. If you can not see any improvements, that probably because of multiple reasons that you need to check one after each other. For example, see which records are giving you error more than the others, and think about some solutions such as designing a new feature engineering method, finding out new features, more data.

This book is also so helpful to guide you through the process of improving. If you do not have time to read the whole book, at least take a look at this article.